Plaintext
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
The Structure of Search Engine Law
James Grimmelmann*
INTRODUCTION ...........................................................................................3!
I.! SEARCH ENGINE TECHNOLOGY AND BUSINESS ...........................................6!
A.! TECHNOLOGY .......................................................................................7!
1.! Indexing .......................................................................................7!
2.! Queries .........................................................................................8!
3.! Results ..........................................................................................9!
4.! Content ......................................................................................11!
B.! BUSINESS ............................................................................................11!
II.! THE STRUCTURE OF SEARCH ENGINE LAW ...............................................15!
A.! USERS’ INTERESTS ...............................................................................17!
1.! Query Privacy .............................................................................17!
2.! Unbiased Results .......................................................................20!
B.! PROVIDERS’ INTERESTS ........................................................................24!
1.! Minimizing Costs .......................................................................24!
2.! Avoiding Unfair Competition ..................................................27!
3.! Prominent Placement in Results ..............................................31!
C.! THIRD PARTIES’ INTERESTS .................................................................33!
1.! Ownership..................................................................................33!
2.! Reputation .................................................................................36!
3.! Privacy ........................................................................................39!
4.! User Virtue.................................................................................41!
D.! SEARCH ENGINES’ INTERESTS ...............................................................44!
1.! Preventing Search Engine Optimization .................................44!
2. ! Preventing Click Fraud .............................................................46!
* Associate Professor of Law, New York Law School. My thanks for their comments to
Jack Balkin, Yochai Benkler, Shyam Balganesh, Aislinn Black, Michael Carroll, Eric Goldman,
Anne Huang, Dan Hunter, David Johnson, Thomas Lee, Beth Noveck, Frank Pasquale, Guy
Pessach, Chris Riley, Steven Wu, Tal Zarsky, and the participants in the workshops where I
presented earlier versions of this Article. After June 1, 2008, this Article is available for reuse
under the Creative Commons Attribution 3.0 United States license, http://
creativecommons.org/licenses/by/3.0/us/. All otherwise-undated web sites in footnotes were
last visited on August 28, 2007.
1
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
2 93 IOWA LAW REVIEW [2007]
3.! Innovation..................................................................................48!
4.! Competition...............................................................................50!
III.! INTERCONNECTIONS IN SEARCH ENGINE LAW ..........................................51!
A.! CLAIMS AGAINST SEARCH ENGINES AS FUNCTIONAL SUBSTITUTES .........52!
B.! THE PROS AND CONS OF DISCLOSURE AND MANDATED RESULTS ............54!
C.! USER PRIVACY CONCERNS IMPLICATE OTHERS’ INTERESTS ....................56!
D.! SEARCH ENGINE RESULTS AS SPEECH ....................................................58!
E.! TRADEMARKS AND SEARCH ENGINES IN CONTEXT .................................60!
V.! CONCLUSION ............................................................................................62!
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 3
INTRODUCTION
Search engines are the new linchpins of the Internet.1 A large and
growing fraction of the Internet’s immense volume of traffic flows through
them. They are librarians, who bring order to the chaotic online
accumulation of information. They are messengers, who bring writers and
readers together. They are critics, who elevate content to prominence or
consign it to obscurity. They are inventors, who devise new technologies and
business models to remake the Internet. And they are spies, who are asked
to carry out investigations with dispatch and discretion.2
Lawyers and the law have taken notice of search engines. Governments
around the world are casting an increasingly skeptical eye on search engines,
questioning whether their actions have always been in the interests of
society. More and more parties are presenting themselves at the courthouse
door with plausible stories of how they have been injured by search engines.
Only a few foresighted legal scholars have recognized the growing
importance of search engines.3
1. See generally JOHN BATTELLE, THE SEARCH: HOW GOOGLE AND ITS RIVALS REWROTE THE
RULES OF BUSINESS AND TRANSFORMED OUR CULTURE (2005) (describing the history and
significance of Internet search); JEAN-NOËL JEANNENEY, GOOGLE AND THE MYTH OF UNIVERSAL
KNOWLEDGE (Teresa Lavender Fagan trans., 2007) (discussing the implications of search for
European cultural heritage); DAVID A. VISE & MARK MALSEED, THE GOOGLE STORY (2005)
(describing Google’s history); IAN H. WITTEN ET AL., WEB DRAGONS: INSIDE THE MYTHS OF
SEARCH ENGINE TECHNOLOGY (2007) (analyzing the role of search engines as the gatekeepers of
information on the web).
2. See, e.g., R. Scott Rappold, Bumbling Bigg City Burglars Got $12K, COLO. SPRINGS
GAZETTE, July 10, 2007, available at http://www.gazette.com/articles/safes_24620___article.
html/ackerman_google.html (describing criminals who Googled for “how to crack a safe” from
a computer at the office they were burglarizing).
3. Many scholars have written about one legal controversy or another that involves a
search engine. Fewer have linked the multiple, interconnected problems that search engines
raise. The essential articles are Niva Elkin-Koren, Let the Crawlers Crawl: On Virtual Gatekeepers and
the Right to Exclude Indexing, 26 DAYTON L. REV. 179 (2001); Urs Gasser, Regulating Search Engines:
Taking Stock and Looking Ahead, 9 YALE J.L. & TECH. 201 (2006), available at http://www.
yjolt.org/old/files/20052006Issue/Spring06-gasser.pdf; Eric Goldman, Search Engine Bias and the
Demise of Search Engine Utopianism, 9 YALE J.L. & TECH. 188 (2006), available at
http://www.yjolt.org/old/files/20052006Issue/spring06-goldman.pdf; Lucas Introna & Helen
Nissenbaum, Shaping the Web: Why the Politics of Search Engines Matters, 16 INFO. SOC’Y 169 (2000);
Frank Pasquale, Rankings, Reductionism, and Responsibility, 54 CLEV. ST. L. REV. 115 (2006); and
Frank Pasquale & Oren Bracha, Federal Search Commission? Access, Fairness, and Accountability in
the Law of Search (2007), available at http://ssrn.com/abstract=1002453. Two unpublished theses
also take a usefully broad view of search. See Michael Zimmer, The Quest for the Perfect Search
Engine: Values, Technical Design, and the Flow of Personal Information in Spheres of Mobility
(2007) (unpublished Ph.D. dissertation, New York University) (on file with the Iowa Law
Review); Alejandro M. Diaz, Through the Google Goggles: Sociopolitical Bias in Search Engine
Design (2005) (unpublished master’s thesis), available at http://epl.scu.edu:16080/
~stsvalues/readings /Diaz_thesis_final.pdf. Joshua A.T. Fairfield’s The Search Interest in Contract,
92 IOWA L. REV. 1237 (2007), is, one may hope, the first of a wave of scholarly attention to the
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
4 93 IOWA LAW REVIEW [2007]
This Article provides a roadmap to the legal issues posed by search. It
indicates what questions we must consider when thinking about search
engines, and it details the interconnections among those questions. It does
not endorse any particular normative framework for search. Nor does it
recommend who should regulate search.4 Instead, it provides an analytic
foundation to distinguish informed decisionmaking from random flailing.
The diverse questions of law it discusses form a coherent set because
each affects the same few information flows. The essence of a search engine
is that it combines its own knowledge of available content with user queries
to provide recommendations to its users; the doctrines and policy values this
Article discusses relate directly to this core process. Other law affects search
engines—Google’s well-publicized Initial Public Offering (“IPO”), for
example, raised substantial issues of securities law,5 and search engines have
been sued for employment discrimination6—but these other issues can be
resolved on their own merits, in isolation. In contrast, the concerns
discussed in this Article must be balanced with one another because each
relates to the same few information flows. Pushing on one affects the others.
Part I explains how modern search engines function and describes the
business environment within which they operate. Search engine operations
can be understood in terms of the information flows among four principal
actors: (1) search engines themselves, (2) their users, (3) information
providers, and (4) third parties with interests in particular content flows
(such as copyright holders and censorious governments). There are, in turn,
four significant information flows: (1) the indexing by which a search
engine learns what content providers are making available, (2) user queries
to the search engine for information about particular topics, (3) the results
returned by the search engine to users, and finally, (4) the content that
providers send to users who have found them through searching. Because so
many major search engines are funded through advertising, this Part also
consequences of having omnipresent, powerful search tools available for legal questions that do
not directly involve search.
Although not scholars, some bloggers have so deeply immersed themselves in the
search world that they have a synoptic perspective on the field. A few blogs are indispensable
reading for anyone interested in search. See John Battelle, SEARCHBLOG,
http://battellemedia.com/; Matt Cutts, GADGETS, GOOGLE, AND SEO,
http://www.mattcutts.com/blog/; Philipp Lenssen et al., GOOGLE BLOGOSCOPED,
http://blogoscoped.com/; SEARCH ENGINE WATCH, http://blog.searchenginewatch.com/
blog/; Danny Sullivan et al., SEARCH ENGINE LAND, http://searchengineland.com/.
4. Cf. Pasquale & Bracha, supra note 3 (comparing institutional forms for search
regulation).
5. See Google, Inc., Registration Statement (Form S-1) (Apr. 29, 2004), available at
http://www.sec.gov/Archives/edgar/data/1288776/000119312504073639/ds1.htm.
6. See Elwell v. Google, Inc., No. 05 Civ. 6487, 2006 U.S. Dist. LEXIS 3114, at *1 (S.D.N.Y.
Jan. 31, 2006).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 5
includes a brief survey of how search engine advertising works and the
distinctive fraud problems confronting search engines and their advertisers.
Part II, the heart of the Article, presents a descriptive analysis of the
legal struggles over search, showing how questions of search policy—many
of which have long been latent in different fields of Internet law—are
increasingly confronting lawyers, courts, and regulators. It describes those
struggles in terms of the legitimate interests that each of these actors brings
to debates over search. Users want high-quality results without too great a
sacrifice of privacy. Content providers want favorable placement in search
results without paying more than their fair share of the costs of supporting
search and without facing unfair competition from search engines. Third
parties want to prevent unauthorized distribution of copyrighted content, to
preserve their own privacy, to protect their reputations, and to preserve what
they see as “user virtue.” And finally, search engines want to preserve their
ability to innovate, to protect themselves from fraud, and to ensure that the
search market remains open to competition. Each entry in this list of
interests has its own associated legal theories; this systematic taxonomy
allows us to recognize how any given legal theory affects the search ecology.
Part III then shows, with five examples, how taking a broad view of
search yields otherwise-unavailable insights into pressing controversies. This
is not to say that the end result must be a body of search-specific law,7 but
only that failing to consider the larger forces at work in search is antithetical
to sensible policymaking. First, the broad, systematic view illustrates how
various claims in search engine disputes can serve as functional substitutes
for each other. Second, it shows that the degree of transparency of the
search process is a highly contested variable, with some concerns pressing
for greater transparency and others pressing for less. Third, it illustrates that
user privacy is a deeply knotty problem and that preserving reasonable user
expectations will involve difficult trade-offs with other interests—including
some of the users’ own. Fourth, it shows that we require a theory of search
engine speech; the most sophisticated theory of search-engine-results-as-
speech so far articulated by a court is too simplistic. And fifth, it illustrates
the richness of debates over search engines’ relationships to providers’
trademarks.
Finally, a brief Conclusion takes note of some of the many open issues
facing search engine law and scholarship.
7. Cf. Frank H. Easterbrook, Cyberspace and the Law of the Horse, 1996 U. CHI. LEGAL F. 207,
207 (arguing that there neither is, nor should be, a distinct body of cyberlaw). Lawrence Lessig
has responded that “more than law alone enables legal values, and law alone cannot guarantee
them,” and has argued that cyberlaw (or “Internet law,” depending on one’s view of the
subject) provides a broader view of law itself. Lawrence Lessig, The Law of the Horse: What
Cyberlaw Might Teach, 113 HARV. L. REV. 501, 549 (1999). I do not make the same claims for
search engine law. Search engines are more important in the consideration of what law should
do than in the consideration of what law is.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
6 93 IOWA LAW REVIEW [2007]
I. SEARCH ENGINE TECHNOLOGY AND BUSINESS
Every major Internet application today is a search engine, contains a
search engine, or depends on a search engine.8 Because search is so
increasingly indistinguishable from other applications, we shouldn’t expect
our definition of “search engine” to differentiate clearly between those
things that are search services and those that aren’t. Instead, let’s start with a
definition that accurately describes the core, paradigm cases: a search
engine is a service that helps its users locate content on the Internet. That’s
what Google, Yahoo!, MSN Live, and Ask.com do: help people find stuff
online. So do their smaller competitors, from AltaVista to Zoohoo.
As anyone who’s played with Google’s ever-expanding list of search
services can attest, search engines help users find more than just web pages.
Google Scholar searches journal articles; Yahoo! Local searches businesses
near the user; the Internet Movie Database searches lists of film casts and
crews; and Like.com searches online sales of clothes and jewelry by color,
shape, and pattern. Thus, it’s better to say that search engines help users
find “content” than to say “pages” or “sites.”9
All of these search engines help users find publicly accessible content,
but others work with specialized sets of content not available to the public.
Thus, LexisNexis, for a fee, allows users to search a large proprietary
database of legal and news documents. Similarly, peer-to-peer file sharing
systems such as Gnutella and Grokster allow users to search content that
typically is accessible only through the peer-to-peer service itself.
Go far enough along these axes (away from the web and away from
publicly accessible content) and you will reach things that are only
marginally recognizable as search engines, according to the definition
above. Google Desktop, for example, is one of several competing tools for
users to search their own computers. Not every legal issue affecting search
applies to these borderline cases. But some issues carry over even here:
Google Desktop has raised privacy concerns resembling those that apply to
plain-vanilla Google Web Search.10 The point is that to the extent a
technology resembles the paradigm case of public web search—and a great
many technologies do—it raises many of the issues described below. Anyone
working with that technology needs to think about how those issues fit
8. The description of search engine operations in this Part draws on John Battelle’s THE
SEARCH, supra note 1. Extended discussion of how search engines work and the business models
of the search industry is available there and in David A. Vise and Mark Malseed’s THE GOOGLE
STORY, supra note 1, and WITTEN ET AL., supra note 1.
9. I say “content” rather than “information” to distinguish it from other significant
information flows connected with search.
10. See Electronic Frontier Foundation, Google Copies Your Hard Drive – Government Smiles in
Anticipation, DEEP LINKS, Feb. 9, 2006, http://www.eff.org/press/archives/2006/02/09
(“‘[Google Desktop] highlights a key privacy problem in the digital age.’” (quoting Cindy
Cohn, Legal Director, Electronic Frontier Foundation)).
GRIMMELMANN_FINAL.DOC 11/21/2007 11:53 AM
THE STRUCTURE OF SEARCH ENGINE LAW 7
together. The taxonomy that follows provides a framework for thinking both
about search engines and about the large penumbra of things that resemble
them.
A. TECHNOLOGY
A search engine in isolation is useless. It becomes valuable only through
interactions with content providers and with users. By aggregating
information about providers’ offerings and presenting it in a useful form, a
search engine can match users with appropriate content providers, to the
benefit of both. This matching, however, can antagonize third parties who
would rather that certain connections not be made.11 Visualizing the
information flows between search engines and these three groups illustrates
how search works.
Search involves four information flows, diagrammed in Figure 1: (1) the
search engine gathers content; (2) a user queries the search engine; (3) the
search engine provides the user with results; (4) the user obtains the
content.
Let’s take up these flows in order.
Figure 1. Information Flows in Search
1. Indexing
While users use a search engine to search for content, a search engine
itself must search out the content it is to recommend to users. It must
therefore work with content providers to learn what they have to offer. With
web search, the process is normally driven by the search engine, which uses
automated software agents—“robots,” “spiders,” or “crawlers”—to explore
the web and find content. It generally does so in the same manner that a
11. As discussed in more detail below, such third parties include copyright holders, targets
of libel, censorious governments, and others. See infra Part II.C.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
8 93 IOWA LAW REVIEW [2007]
user would, requesting web pages from content providers’ servers and seeing
what those pages contain.12
Other forms of search involve different forms of information gathering.
Some search engines simply take existing collections of information and
organize them more effectively—a nationwide phone-number search
aggregates information that was, at one time, only available in a shelf of
phone books. Others rely on content providers to come to them. Under
“paid search inclusion,” once practiced by a number of search engines,
content providers pay a search engine and supply it with content; the engine
promises to index any content for the appropriate fee.13 Major search
engines now offer tools for content providers to describe and organize their
content in standardized formats.14
The line between search engine and content provider can be indistinct.
Online merchants, such as Amazon.com, typically provide search engines for
their own sites, as do most sites that aggregate user-supplied content (e.g.,
Wikipedia). One does not have to use the site-specific search engine to
reach the content, but one can. Decentralized peer-to-peer systems use the
same computers both to provide files and to index them.
2. Queries
Content is only one of two inputs to search. The other is the “search
query,” a request by a user for information on a particular topic. Most search
engines use queries made up of a few keywords or short phrases. Some non-
textual search engines allow users to issue queries in more exotic forms,
such as hummed tunes15 or pictures.16 In whatever form, the user provides
the search engine with some criteria to narrow the vast universe of possible
results.
A query is typically only an approximation of the user’s intentions.
Common intentions include navigational queries (the user wishes to find a
specific site or datum), informational queries (the user wishes to find
information on a topic), and transactional queries (the user wishes to
12. See Google, How Google Crawls My Site, http://www.google.com/support/
webmasters/bin/topic.py?topic=8843.
13. See Danny Sullivan, The Evolution of Paid Inclusion, SEARCH ENGINE WATCH, July 2, 2001,
http://searchenginewatch.com/showPage.html?page=2163971 (“Paid inclusion programs
mean that, in exchange for a payment, a search engine will guarantee to list pages from a web
site.”).
14. See, e.g., Google, About Google Base, http://base.google.com/support/bin/
answer.py?answer=59260&hl=en (describing Google Base as “a place where you can easily
submit all types of online and offline content, which we’ll make searchable on Google”).
15. See Midomi Video Tour, http://www.midomi.com/index.php?action=main.video
(“midomi lets you find music by singing or humming!”).
16. See Bob Tedeschi, Shopping Site Offers a Way to Raid a Celebrity’s Closet, N.Y. TIMES, Nov.
13, 2006, at C4; Like Visual Search, http://www.like.com (“Like.com is a visual search engine
that lets you find items by color, shape, and pattern”).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 9
perform an activity, such as purchasing a good).17 Because words have
multiple meanings, the same query reasonably could be directed at many
different possibilities, even within one of these categories. Further, the user
may not have in her head a clear idea of what she is searching for. What she
is searching for may not even exist. By way of example, consider a search for
“apple.” The user might have intended any of the following:
!To find the home page for Apple Computer (navigational);
!To learn about apples, the fruit (informational);
!To purchase an Apple MacBook computer (transactional);
!To purchase apples, the fruit (transactional, but different);
!To learn about oranges (informational, but confused);
!To find the home page for Apple Records, the Beatles’ record
label (navigational, but no such page exists18); or
!To test whether her connection to the Internet is working.
Along with the query itself may come various user information, such as
geographic location, preferred types of results, operating system and
browser, and preferences among search results revealed through clicks on
past search results. Some search engines, such as Google and Yahoo!, keep
extensive histories on the searches and preferences of registered users.
Personalized search engines may customize their results, showing different
results to users with different geographic locations or announced interests.
3. Results
In the defining step of search, the search engine returns to the user
information about content relevant to her query. Web search engines
typically show results on a web page that lists results ten at a time, starting
with those results it thinks she would probably most like to see and following
in descending order with results it thinks are probably less relevant. Each
search result normally contains the name of the identified piece of content,
its location, and a very short summary or excerpt that shows how the content
relates to the query. As the user inspects a set of search results, she may
choose to refine her query, choosing slightly different keywords in an
attempt to better convey her intention to the search engine. The engine, in
turn, will supply her a different set of results. This process is iterative, and
17. See Andrei Broder, A Taxonomy of Web Search, SIGIR FORUM, Fall 2002, available at
http://www.sigir.org/forum/F2002/broder.pdf. Broder’s taxonomy is obviously not exhaustive,
nor are its three categories entirely distinct from one another, but it provides a good first
approximation.
18. The closest substitutes are probably the home page for Apple Corps, Ltd.,
http://www.thebeatles.com/ (the conglomerate parent of Apple Records), and The Complete
Apple Records, http://www.schomakers.com/ (an unaffiliated site with a complete Apple Records
discography).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
10 93 IOWA LAW REVIEW [2007]
thus, even within a single search session, query and results flows may repeat
any number of times.19
This is the step in which search engines really differentiate themselves:
by using varied algorithms to summarize and organize the vast seas of
available content. At one time, web search engines simply scanned the text
of web pages to determine which topics the pages discussed. They then
augmented this technique by analyzing additional information about pages
(called “metadata”), such as their age, the number of links they contain, or
the keywords used by their authors to describe them. More recently,
powerful link-structure techniques involving study of which web pages link
to which other pages have become the dominant web search paradigm.20
The biggest technical challenge in search today is integrating these
computationally expensive analyses with a user’s particular query. A user’s
query must be answered in a few seconds, but it can take a server farm of
many thousands of computers weeks to build an index of the web. Search
engines that deal with small domains can afford to gather information from
content providers in response to a particular query, but most search engines
preprocess information about available content to produce a specially
optimized index that can be consulted quickly. The most extreme form of
this optimization is to have pre-generated lists of results and merely allow
users to choose, in effect, from a predetermined list of queries. Such search
engines function like the index of a book, allowing a user to find quickly
anything the search engine has specifically chosen to show.
Search technologists also distinguish between query-dependent and
query-independent indexes. Google’s PageRank, for example, is an example
of the latter; PageRank estimates the general popularity of a web page.
Google uses PageRank to help sort its lists of results relevant to a particular
query and return more popular pages first, but PageRank does not by itself
say anything about which pages are relevant. That is the job of query-
dependent analysis, such as looking in the pages for query terms and their
synonyms. Other search engines have their own algorithms, but this
integration of query-dependent and query-independent information is a
standard technique.
Search engines are also increasingly learning from the large volumes of
query data they have accumulated. A user’s history of queries can provide
useful information about her probable intentions—for example, whether
she tends toward navigational or transactional queries. Similarly, search
19. See Goldman, supra note 3, at 196. As Goldman explains, this refinement of results is a
critical process by which users correct for ambiguities in their initial queries—and the
interactivity of the process argues against trying to infer a single fixed meaning for a query. Id.
20. The best available reference on search algorithms is AMY N. LANGVILLE & CARL D.
MEYER, GOOGLE’S PAGERANK AND BEYOND: THE SCIENCE OF SEARCH ENGINE RANKINGS (2006),
which provides a clear presentation of Google’s PageRank and related algorithms and contains
an extensive bibliography of the technical literature on search ranking.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 11
engines gain useful feedback into their own successes and failures by seeing
which results users click on or by noticing long strings of searches on related
terms, which may indicate that the user is having trouble finding what she’s
looking for.
4. Content
In the end, the user cares about the content. Most often, she takes the
location information given her by the search engine and uses it to approach
the appropriate content provider. For web pages, that typically means she
reads the page and consumes the information on it. For other goods and
services, she may purchase them (or decide not to, after considering her
options). Typically, both the user and content provider value being matched
up, since this sort of exchange is why the content provider is online and why
the user turned to a search engine.
Search engines themselves provide some content to users. Simply telling
users what the content is provides some information about it; often, a web
page title, a short excerpt of its text, or a thumbnail image may be enough to
meet the user’s needs. Some search engines cache content, storing copies to
make it easier for users to receive it quickly. Others archive content,
enabling users to receive it even when the original provider cannot be
reached.
B. BUSINESS
It is impossible to understand the legal controversies over search
without some understanding of the most common search business models.
The overwhelmingly predominant model for web search today is contextual
advertising, in which the search engine shows its users advertisements
alongside the search results. Most commonly, these ads are brief two- or
three-line blocks of text containing a hyperlink to the web site being
advertised and various visual indicators marking them as advertisements.21
These ads are still often search “results” in the sense that the search engine
presents particular ads based on the user’s query.22 Search engine
21. But see DEBORAH FALLOWS, PEW INTERNET & AM. LIFE PROJECT, SEARCH ENGINE USERS
17 (2005), http://www.pewinternet.org/pdfs/PIP_Searchengine_users.pdf (finding that sixty-
two percent of search engine users were unaware of the distinction between search results and
advertisements).
22. See BATELLE, supra note 1, at 109–21. See generally Benjamin Edelman & Michael
Ostrovsky, Strategic Bidder Behavior in Sponsored Search Auctions, 43 DECISION SUPPORT SYS. 192
(2007), available at http://www.benedelman.org/publications/cycling-060703.pdf (discussing
ways to reduce strategic behavior). See also F. Gregory Lastowka, Search Engines Under Siege: Do
Paid Placement Listings Infringe Trademarks?, 14 INTELL. PROP. & TECH. L.J. 1, 2 (2002). The
process is closely related to paid inclusion; in each, a content provider has paid the search
engine in the hope of being seen by users entering particular queries. Id. at 2. But see Thomas A.
Weber & Zhiqiang (Eric) Zheng, A Model of Search Intermediaries and Paid Referrals, (Wharton
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
12 93 IOWA LAW REVIEW [2007]
developers have created sophisticated bidding algorithms that balance
advertisers’ willingness to pay with the popularity of their ads in choosing
which so-called “sponsored links” to show.23 Thus, content providers can be
found by users either by being listed as search results or by advertising on
search engines.
Search engines use three common billing techniques to sell ads. An
advertiser using pay-per-impression pays a given fee to the engine each time
a user sees the ad. Under pay-per-click, the advertiser pays each time a user
not only sees the ad but also clicks on it. Under pay-per-conversion (also
known as pay-per-action or pay-per-performance), the advertiser pays only
when the user makes a purchase or takes some similar action that indicates
serious interest in the advertiser’s site. Pay-per-conversion is the most closely
correlated with actual sales, but it requires that advertisers turn over
significant information to the search engine so that billing can be properly
calculated. Pay-per-click currently strikes the most popular overall
compromise between accuracy and convenience.24
A few web search engines do not show ads. Providers with site-specific
search engines can monetize them directly because every result will be from
their own site. Some search engines are maintained out of altruism—the
search facility is provided ad-free as a public service. A few search engines
have attempted to make money by analyzing user search patterns and selling
aggregate information.25
Search advertising has generated its own distinctive forms of fraud. In
click fraud, users upset at a particular advertiser or in competition with it
Sch. OPIM, Working Paper No. 02-12-01, 2006), available at http://ssrn.com/abstract=601903
(arguing that ranking based on provider bids decreases overall social welfare).
23. See BATTELLE, supra note 1, at 142.
24. A full discussion of the business issues involved in search engine advertising is beyond
the scope of this paper. As Tal Zarsky has explained in correspondence, the different billing
models have significantly different implications for the incentives of search engines and
advertisers and for the privacy of search engine users. Thus, for example, under a pay-per-click
system, search engines will try to favor not just those ads for which advertisers will pay the most
per click but also those ads that will generate the most clicks. Advertisers, in turn, will therefore
be able to purchase prominent advertising placement more cheaply if their ads are well
designed to encourage clicks. The process of monitoring clicks, however, is both vulnerable to
fraud and requires close observation of user behavior—two facts that implicate the general
tension between openness and transparency in search advertising and search engine
operations. E-mail from Dr. Tal Zarsky, Law Faculty, Haifa University, to James Grimmelmann
(Dec. 31, 2006, 12:56) (on file with the Iowa Law Review); see also Ben Elgin, The Vanishing Click-
Fraud Case, BUSINESSWEEK ONLINE, Dec. 4, 2006, available at http://www.businessweek.
com/print/technology/content/dec2006/tc20061204_923336.htm.
25. The Alexa Toolbar uses this strategy. Once installed in a user’s browser, it displays
extra information about the pages that the user visits, including related links. In exchange, it
can track which pages users visit. This aggregate user data is valuable even just as a Nielsen-type
rating for web pages.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 13
repeatedly view and click on its ads to run up the advertiser’s bills.26 When
search engines use their ad-serving infrastructure to act as advertising
brokers and place ads on other web sites, affiliate fraud becomes a
possibility; there, web sites sign up as affiliates to have ads placed on their
pages and then turn around and click on the ads themselves. They pocket
some of the money, while the bill goes to the advertisers.27
The great demand for high placement28—when combined with the
zero-sum nature of ranking decisions—leads to search engine optimization
(“SEO”): the business of redesigning content (or creating it) to attract
search engines and convince them to rank content highly.29 Some “white
hat” SEO techniques are generally considered desirable because they make
the content easier for search engines and users to access. Other, “black hat”
techniques involve mimicking the superficial features that search engines
use as proxies for quality content. When search engines scanned page text
and keywords, optimizers would hide popular keywords in invisible, tiny text
on a page or show a search engine a different page (one larded with
thousands of keywords) than the one shown to users. As search engines
shifted to analyzing link structure, optimizers switched to creating “link
farms”: sets of thousands of sites and pages pointing to each other,
mimicking a community of real users and hoping to trick search engines
into treating them as authoritative, popular sources of information.30 Search
engines and black-hat SEOs are locked in a technical arms race that pits
increasingly sophisticated algorithms that distinguish fraudulent from
26. See Charles C. Mann, How Click-Fraud Could Swallow the Internet, WIRED, Jan. 2006,
available at http://www.wired.com/wired/archive/14.01/fraud_pr.html (discussing the
ramifications of click fraud).
27. Affiliate advertising and affiliate fraud do not involve search as such. They are,
however, a good example of the Protean nature of search technology. The algorithms that can
determine whether a web page is relevant to a particular query are the same kind of algorithms
that can determine whether a web page is a good place to show a particular ad.
28. See Nico Brooks, The Atlas Rank Report: How Search Engine Rank Impacts Traffic,
ATLAS INST. (2004), http://www.atlassolutions.com/WorkArea/showcontent.aspx?id=1440&
LangType=133. Users are far more likely to click on the first result in a list shown to them than
on any other result; if they are shown a page of ten results, only a small fraction of them will
click through to see even the second page. Results after the hundredth or so may as well not
exist.
29. See David Kesmodel, ‘Optimize’ Rankings at Your Own Risk, WALL ST. J. ONLINE, Sept. 23,
2005, http://www.startupjournal.com/ecommerce/20050923-kesmodel.html; Google, What’s
an SEO? Does Google Recommend Working with Companies That Offer to Make My Site
Google-Friendly?, http://www.google.com/support/webmasters/bin/answer.py?answer=35291.
30. Link farms have cropped up in some surprising places. See Andy Baio, Wordpress
Website’s Search Engine Spam, WAXY.ORG, Mar. 30, 2005, http://www.waxy.org/archive/
2005/03/30/wordpres.shtml (homepage for “popular open-source blogging software”); Philipp
Lenssen, Forbes Spam?, GOOGLE BLOGOSCOPED, Apr. 17, 2007, http://blogoscoped.
com/archive/2007-04-17-n53.html (Forbes magazine); Blake Ross, Stanford Daily Link Spam
Harms the Web and Students, BLAKEROSS.COM, May 27, 2005, http://blakeross.
com/index.php?p=136 (college newspapers).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
14 93 IOWA LAW REVIEW [2007]
authentic content against increasingly subtle forms of mimicry.31 As might
be expected, black-hat-SEO techniques are highly controversial, and the line
between black- and white-hat techniques is both unclear and contested, as is
the line between authentic and fraudulent content.
Some of the business practices of those who make their living in the
search ecosystem have more broadly harmful consequences. On the one
hand, search engines provide one of the main sources of traffic for various
fraudulent schemes—unsuspecting users landing on pages they don’t really
want to see can be shown advertisements, tricked into downloading
spyware,32 or scammed out of their money.33 In order to acquire that search
traffic in the first place, black-hat SEOs use all sorts of techniques that make
the Internet less usable. They have registered domains fraudulently,34 posted
link-filled comments to blogs and discussion boards,35 created fake web sites
and blogs,36 hijacked popular domains,37 and sent hyperlinks in e-mail,
instant messages, and even in requests for web pages.38 Major search engines
provide a centralized target that makes such schemes profitable.
31. See, e.g., Lee Gomes, Our Columnist Creates Web ‘Original Content’ But Is in for a Surprise,
WALL ST. J., Mar. 1, 2006, at B1, available at http://online.wsj.com/public/article/
SB114116587424585798-0qH9qUYuUug__vRSFKGvxIEwLGw_20070301.html?mod=blogs;
Wikipedia, Wikipedia: Send in the Clones, http://en.wikipedia.org/wiki/Wikipedia:
Send_in_the_clones (noting that much Wikipedia content is now available on other sites and
discussing potential methods of competition). Cf. Stephen Baker, Asbestos and the Art of Blogging
for Money, BUSINESSWEEK ONLINE, May 27, 2005, http://www.businessweek.com/
the_thread/blogspotting/archives/2005/05/asbestos_and_th.html (describing an experiment
in which Michael Buffington created a topical blog on asbestos specifically to capture some of
the money that lawyers would pay per click on ads triggered by related keywords). There is no
clear line between such experiments and “real” professional blogs.
32. See Ben Edelman & Hannah Rosenbaum, The Safety of Internet Search Engines, MCAFEE
SITEADVISER, May 12, 2006, http://www.siteadvisor.com/studies/search_safety_may2006.html
(“Despite search engines’ efforts, we see too many sites trying to deceive unsuspecting users.”).
33. See Ben Edelman, False and Deceptive Pay-Per-Click Ads, Oct. 10, 2006,
http://www.benedelman.org/ppc-scams/ (describing ads that make false or misleading
claims).
34. See, e.g., Declan McCullagh, Dotster Named in Massive Cybersquatting Suit, CNET NEWS,
June 2, 2006, http://news.com.com/2102-1032_3-6079567.html (discussing a lawsuit against a
large domain-name registrar).
35. See Six Apart, Six Apart Guide to Combating Comment Spam, http://www.sixapart.
com/pronet/comment_spam.html.
36. See Christopher Heun, Spam Blogs Pollute Internet Search, INFORMATIONWEEK, May 15,
2006, http://www.informationweek.com/showArticle.jhtml?articleID=187202310.
37. See Will Baude, The Remains of Crescat, CRESCAT SENTENTIA (NOV. 5, 2006), http://www.
crescatsententia.net/archives/2006/11/05/#006824 (describing the loss of
“crescatsententia.com,” previous home of a popular legal blog, to an SEO domain-squatter).
38. See Referer Log Spam, METAFILTER (Oct. 24, 2002),
http://www.metafilter.com/21063/ (discussing SEO attempts to fill a referer log with spammed
URLs).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 15
II. THE STRUCTURE OF SEARCH ENGINE LAW
The set of laws potentially applicable to search may seem bewilderingly
large. It has, however, a recurring deep structure that becomes evident if we
focus on four concepts: the actors involved, the information flows among
them, the interests that they bring to search, and the legal theories that they
use to vindicate their interests. We have already met the key actors: (1)
search engines, (2) content providers, (3) users, and (4) concerned third
parties. We have also discussed the relevant information flows: (1) indexing,
(2) queries, (3) results, and (4) content. This Part details these actors’
interests in these information flows and examines the legal theories
associated with those interests.
Every dispute involving a search engine is, at base, a dispute among
these constituencies. To see what is really at stake in any given legal battle
over search, it is useful to ask what these groups stand to gain or lose.
On the one hand, search engines create enormous social benefits. They
allow willing users and content providers to find each other, reducing
transaction costs and enabling mutually beneficial exchanges. These benefits
depend on the contributions of users, providers, and search engines in the
form of queries, content, and ranking algorithms, respectively. Good search
engine policy would therefore give each group appropriate incentives to
maximize its productive contributions while deterring rent-seeking behavior.
On the other hand, search engines can also cause enormous harms to
particular parties. By controlling the process matching users and content
providers, they create winners and losers within these communities. Both
users and providers entrust search engines with valuable information and
may be upset at the terms on which search engines reveal that information.
Third parties who would prefer that certain content not flow from providers
to users also are injured when search engines enable such flows. Good
search engine policy will prevent search engines from unreasonably
inflicting serious harms on others.
Complicating matters, these two features of search are inextricably
intertwined. Search engines do not generally cause harms out of inherent
malice. They cause harms in the process of serving their other
constituencies. It is precisely the fact that search engines create enormous
value that gives them such power to cause enormous harm. Users value
search engines because the search engines help them pick and choose
among possible providers; third parties are most upset at new content flows
precisely when users and providers value those flows the most. Attempts to
remediate particular harms, then, almost invariably involve a contest among
the interests of these other constituencies. The law’s choice to intervene or
not is a choice among their interests.
The central position that search engines occupy also creates problems
even when the balance between their various constituencies seems
appropriate. If the balance comes with too much deference to search
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
16 93 IOWA LAW REVIEW [2007]
engines, the risk is that they will behave unaccountably and upset the
balance by aligning themselves with one group against another. But if the
balance comes with too many restrictions on the actions of search engines,
the risk is that those restrictions will squander the innovative potential of
search engines.
With these general principles in mind, let us take up the specific
interests that each constituency brings to any discussion of search law:
Users are the most obvious group served by search engines. When they
make search queries, they reveal some potentially private information; thus,
they may be harmed if a search engine misuses or reveals that information.
They also desire useful search results, so a search engine can harm them by
providing low-quality or deliberately biased results.
Providers have three interlocking interests in search. On the one hand,
they usually value prominent placement in search results because of the
many users they can draw in. On the other hand, they’d prefer not to
shoulder more than what they perceive as their fair share of the costs
involved in providing search. That means limiting bandwidth and server
time they burn to support indexing by search engines; it also means limiting
the search advertising they buy. Content providers also fear that search
engines will compete unfairly with them, either by opening up new ways to
view content or sometimes simply by delivering content directly to users.
Third parties sometimes fear the content flows that search engines
enable, for a variety of reasons. Some have a copyright interest in the
material now flowing freely; some may be the subjects of that material and
feel that it defames them or invades their privacy; some (most often
governments) may simply wish to censor the flow of content they consider
objectively harmful.
And finally, search engines themselves want to keep their business
models viable. That means dealing with the operational threats of SEO and
click fraud, preserving the incentives to innovate in developing new forms of
search technology, and maintaining competitive freedom from the risk that
other search engines will unfairly dominate the market for search.
These controversies are summarized in Table 1:
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 17
Table 1. Interests in Search
Constituency Interest Information Sample Legal
Flows Theories
Users Privacy Queries Information
Privacy,
Contract
Quality Results Consumer
Protection
Providers Costs Indexing, Trespass,
content Contract
Unfair Content Copyright,
Competition Trademark,
Contract
Placement Results Trademark,
Business torts
Third Parties Ownership Content Copyright,
Trademark
Reputation Content Defamation
Privacy Content Information
Privacy
User Virtue Content Direct
Regulation
Search SEO Indexing, Fraud
Engines Results
Click Fraud Results Contract
Innovation All Intellectual
Property
Competition All Antitrust
The remainder of this Part takes up these issues in order.
A. USERS’ INTERESTS
1. Query Privacy
Effective search requires that users disclose information about their
interests and intentions. Whether they click on a topic heading in an index
or craft a complex query with various exclusions and inclusions, the very fact
that they are curious about something will be evident in their queries. If they
are repeat users, the search engine may be able to construct an extensive
history of their queries. It may also be able to correlate this curiosity with
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
18 93 IOWA LAW REVIEW [2007]
users’ actual behavior in obtaining content or with other information it has
about them from their use of other applications and features it provides. In
addition to collecting in-depth data on each user, a search engine also has
broad access to information about many users.
Much of this data is personal or sensitive. When AOL publicly released
the search queries of some 650,000 search users, the logs included queries
such as “can you adopt after a suicide attempt,”39 “cocaine in urine,”40 and
“How to deal with mental abuse in a Christian marriage.”41 Even though the
search logs identified users only by pseudonymous numbers, reporters
showed by example that it was possible to take a list of searches and identify
the searcher; people have a tendency to search on their own names, their
addresses, and other personally identifying details.42 Given the sensitivity of
this information and the ease of linking it back to particular individuals,
users have an evident privacy interest that their queries not be misused. Even
where the information cannot be linked to an individual, it can still be used
in ways that cause privacy harms.43
Case law dealing with collection of personal information online suggests
that users will not enjoy significant legal recourse against search engines for
the misuse of their queries. Multiple courts have held that users fail to state a
claim when they allege that web advertising services (with the cooperation of
the web sites on which the ads appear) have captured their browsing habits
and tracked them through time.44 Once in possession of the information, an
engine is free to disclose it to others, unless it has undertaken not to.45 A
39. Declan McCullagh, AOL’s Disturbing Glimpse into Users’ Lives, CNET NEWS, Aug. 8, 2006,
http://news.com.com/2100-9588_22-6103098.html.
40. Id.
41. Lee Gomes, What Are Web Surfers Seeking? Well, It’s Just What You Think, WALL ST. J., Aug.
16, 2006, at B1.
42. Michael Barbaro & Tom Zeller Jr., A Face is Exposed for AOL Searcher No. 4417749, N.Y.
TIMES, Aug. 9, 2006, at A1.
43. In the terminology of Daniel Solove’s A Taxonomy of Privacy, 154 U. PA. L. REV. 477,
491–561 (2006), a search engine could harm a user through disclosure, breach of
confidentiality, decisional interference, exposure, identification, secondary use, aggregation,
and surveillance. For extensive discussion of the privacy implications of search, see Zimmer,
supra note 3. Tal Zarsky, in Mine Your Own Business!: Making the Case for the Implications of the Data
Mining of Personal Information in the Form of Public Opinion, 5 YALE J.L. & TECH. 4, 37–38 (2003),
argues that the aggregation of information about consumer habits can lead to a harmful
“autonomy trap.” Julie Cohen’s A Right to Read Anonymously: A Closer Look at “Copyright
Management” in Cyberspace, 28 CONN. L. REV. 981 (1996) explains clearly the dangerous chilling
effect of surveillance of one’s reading habits.
44. See, e.g., Chance v. Ave. A, Inc., 165 F. Supp. 2d 1153, 1155 (W.D. Wash. 2001)
(interpreting 18 U.S.C. §§ 1030, 2511(2)(a), 2701 (2000)); In re DoubleClick Inc. Privacy Litig.,
154 F. Supp. 2d 497, 500 (S.D.N.Y. 2001) (interpreting 18 U.S.C. §§ 1030, 2510, 2701).
45. See, e.g., In re JetBlue Airways Corp. Privacy Litig., 379 F. Supp. 2d 299, 330 (E.D.N.Y.
2005) (interpreting 18 U.S.C. § 2701). Legislation to change this baseline and require the
deletion of stored queries was introduced in the 109th Congress, but it died in committee. See
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 19
search engine with carefully drafted terms of service, therefore, can largely
immunize itself from user privacy suits. Even a violation of its own stated
privacy policy may not expose it to significant legal risk.46
Even though courts may not obligate a search engine to keep queries
private, is it permitted to do so against the legal demands of others? Answers
vary. The Fourth Amendment and various statutes allow several procedures
by which computers can be searched, balancing the government’s showing
of relevance to an investigation against user expectations of privacy.47 A full
search warrant, properly supported by an affidavit showing probable cause,
will trump any expectation of privacy. Internationally, search engine
operators (albeit not in their roles as search engines) have shown a
willingness to comply with government demands for identifying data, even
when the consequences for the identified users are severe.48 Some have been
proactive in cooperating with law enforcement.49
Where the demand comes from a private-sector third party, users may
have more leverage. A traditional subpoena duces tecum issued to an
intermediary, such as a search engine, affords the intermediary and the
adversary in the underlying litigation an opportunity to object.50 The court
will balance the relevance of the information to the litigation against the
burden on the recipient of the subpoena. Google recently resisted a
Department of Justice subpoena for a random sample of user queries; it
successfully argued that users’ loss of trust in a search engine that releases
their queries would constitute a burden.51 Courts considering the use of
subpoenas to learn the identity of particular users have developed tests that
depend on the speech interests of the users,52 although they have yet to
Eliminate Warehousing of Consumer Internet Data Act of 2006, H.R. 4731, 109th Cong.
(introduced Feb. 8, 2006, referred to committee Feb. 17, 2006).
46. See, e.g., In re Nw. Airlines Privacy Litig., No. 04-126, 2004 U.S. Dist. LEXIS 10580, at
*15–17 (D. Minn. June 6, 2004) (dismissing a breach of contract claim); In re Geocities, 127
F.T.C. 94, 94 (1999) (entering an FTC consent order without fine or punishment).
47. See generally COMPUTER CRIME AND INTELLECTUAL PROP. SECTION, CRIMINAL DIV., U.S.
DEP’T OF JUSTICE, SEARCHING AND SEIZING COMPUTERS AND OBTAINING ELECTRONIC EVIDENCE IN
CRIMINAL INVESTIGATIONS (2002), available at http://www.cybersafe.gov/criminal/cybercrime/
s&smanual2002.pdf.
48. See, e.g., Anne Broache, Google to Hand over Brazilian User Data, CNET NEWS, Sept. 5,
2006, http://news.com.com/2061-10812_3-6112176.html; Philippe Naughton, Yahoo Blamed for
Jailing of Chinese Reporter, TIMES (LONDON) ONLINE, Sept. 7, 2005, http://business.
timesonline.co.uk/tol/business/industry_sectors/media/article563772.ece.
49. See, e.g., Jonah Engle, Buyer Beware: eBay Security Chief Turns Website into Arm of the Law,
THE NATION ONLINE, June 20, 2003, http://www.thenation.com/doc/20030707/engle.
50. See FED. R. CIV. P. 45(c).
51. Gonzales v. Google, 234 F.R.D. 674, 683–86 (N.D. Cal. 2006). AOL, Yahoo!, and
Microsoft did not resist subpoenas served on them for similar data. Id. at 679.
52. See, e.g., O’Grady v. Superior Court, 139 Cal. App. 4th 1423 (Cal. Ct. App. 2006)
(granting reporters’ motion to quash a subpoena served on an ISP to learn the identity of a
confidential source); Sony Music Entm’t Inc. v. Does 1–40, 326 F. Supp. 2d 556, 568 (S.D.N.Y.
2004) (requiring an ISP to disclose the identities of copyright-infringement defendants); Doe v.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
20 93 IOWA LAW REVIEW [2007]
determine what speech interests users have in anonymous search. Some
courts have held that the statutory subpoena process of the Digital
Millennium Copyright Act53 does not apply to Internet Service Providers
(“ISPs”) that do not themselves host allegedly infringing files,54 using
reasoning that would appear to protect search engines as well.
One further possibility for protecting users’ interest in privacy is self-
help. Users might modify their searching behavior to prevent their queries
from being correlated or linked back to them. Possible schemes involve
anonymizing their connection to the search engine,55 querying multiple
search engines with partial queries and then correlating results,56 and
sending a cloud of off-topic queries to mask the true query of interest.57 The
practical effectiveness of these techniques remains questionable.58
2. Unbiased Results
Users turn to search engines to help them find useful, high-quality
information. They therefore share search engines’ interest in defeating SEO
and promoting fair competition.59 They want search engines to actually give
them the highest-quality results that the engines are capable of delivering.
Users are arguably harmed when a search engine gives some users higher-
quality results than others or when it favors one content provider over
another for reasons with which the users would disagree. Both could
constitute bias, which Batya Friedman and Helen Nissenbaum have defined
in this context as “systematic[] and unfair[] discriminat[ion] against certain
individuals or groups of individuals in favor of others.”60
2TheMart.com, Inc. 140 F. Supp. 2d 1088, 1097–98 (W.D. Wash. 2001) (requiring a showing
that the need to learn the identity of a non-party is “directly and materially relevant to a core
defense in the underlying . . . litigation”).
53. Digital Millennium Copyright Act, 17 U.S.C. § 512(h) (2000).
54. In re Charter Commc’ns, Inc., Subpoena Enforcement Matter, 393 F.3d 771, 777–78
(8th Cir. 2005); Recording Indus. Ass’n of Am., Inc. v. Verizon Internet Servs., Inc., 351 F.3d
1229, 1238–39 (D.C. Cir. 2003).
55. See Ryan Singel, How to Foil Search Engine Snoops, WIRED.COM, Jan. 20, 2006,
http://www.wired.com/news/technology/1,70051-0.html (recommending that users delete
“cookies” each week or use a service that “masks the origins” of a query and evades filters).
56. See generally Benny Chor et al., Private Information Retrieval, 45 J. OF THE ACM 965
(1998), available at http://www.cs.technion.ac.il/~benny/PIR.pdf.
57. See Wakaha Ogata & Kaoru Kurosawa, Oblivious Keyword Search, 20 J. COMPLEXITY 356,
357 (2004); Daniel C. Howe & Helen Nissenbaum, TrackMeNot, http://mrl.
nyu.edu/~dhowe/TrackMeNot/ (describing a program that obfuscates actual web searches).
58. See, e.g., Bruce Schneier, TrackMeNot, SCHNEIER ON SECURITY (Aug. 23, 2006),
http://www.schneier.com/blog/archives/2006/08/trackmenot_1.html (“Let’s count the ways
[TrackMeNot] doesn’t work.”).
59. See infra Part II.D.
60. Batya Friedman & Helen Nissenbaum, Bias in Computer Systems, 14 ACM TRANSACTIONS
ON INFO. SYS. 330, 332 (1996) (emphasis omitted).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 21
The first significant challenge here is to distinguish bias from simple
failure by the search engine to do as well as it could have. Several problems
make it difficult to set a proper baseline of “unbiased” results. As noted
above, different users may have different needs and desires, a given query
may reflect any number of different intentions, and even the user may not
know what she is searching for when she queries a search engine. Some have
even questioned whether users are the proper judges of what search results
would be best for them.61
Even without a precise definition of “unbiased” results, some kinds of
search decisions are troubling. Direct censorship—removing content from a
search engine’s index at the demand of a third party—is offensive to free-
speech values.62 In China, for example, at the demand of the government,
major search engines try to prevent users from finding information on the
banned Falun Gong movement.63 A search engine could also bias its results
by ranking favored content more highly and disfavored content more
poorly; the disfavored content can still be found, but only if you’re
specifically looking for it. Both liberal and conservative groups have accused
Google of bias toward the other in its advertising policies.64 The concern is
commercial as well as political: some have claimed (albeit without direct
evidence) that search engines systematically favor their own advertisers and
affiliated corporate providers.65
Technical design features of search engines can also introduce
unconscious structural biases in their coverage and ranking of content.66
John Hiler has suggested that search engines may rely too heavily on
61. See CASS SUNSTEIN, REPUBLIC.COM 8–10 (2001) (arguing that users will choose to see
only content confirming their preexisting biases if they are given total control over their
information inputs). But see NICHOLAS NEGROPONTE, BEING DIGITAL 163–65 (1995) (positing
that users have tastes both for narrowly personalized information and for the common
information seen by many others).
62. See infra Part II.C, and particularly Part II.C.4, for more on the issues raised by search
censorship.
63. See Clive Thompson, Google’s China Problem (and China’s Google Problem), N.Y. TIMES,
Apr. 23, 2006, (Magazine), at 63, available at http://www.nytimes.com/2006/04/23/magazine/
23google.html.
64. See, e.g., Google’s Gag Order: An Internet Giant Threatens Free Speech, PERRSPECTIVES, June
20, 2004, http://www.perrspectives.com/articles/art_gagorder01.htm (claiming apparent
conservative bias); Rightmarch.com, Google says NO to Conservative Ads!, http://www.
rightmarch.com/google.htm (claiming liberal bias). But see Eric Ulken, A Question of Balance:
Are Google News Search Results Politically Biased? (May 5, 2005) (unpublished report, USC
Annenberg School for Communication), available at http://ulken.com/thesis/googlenews-bias-
study.pdf (claiming no bias toward either side in Google News selection of articles).
65. See, e.g., Sergey Brin & Lawrence Page, The Anatomy of a Large-Scale Hypertextual Web
Search Engine, at App. A, http://infolab.stanford.edu/~backrub/google.html (“[W]e expect that
advertising funded search engines will be inherently biased towards the advertisers and away
from the needs of the consumers.”). But see Google, Why We Sell Advertising, Not Search
Results, http://www.google.com/honestresults.html.
66. See Introna & Nissenbaum, supra note 3, at 175–76.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
22 93 IOWA LAW REVIEW [2007]
weblogs because weblogs tend to be frequently updated and contain many
hyperlinks.67 Studies of relative traffic and links to web sites have also caused
some to discern a “Googlearchy,” in which the most popular content
receives more attention from users and therefore becomes even more
popular, creating a popularity-feedback loop that new providers can never
hope to break into.68 Mathematical models and empirical studies both
support and undercut this theory.69 Separating cause from effect in the
wildly uneven popularity of content has proven difficult, as have attempts to
show whether these differences are undesirable or not.70
If users or regulators are concerned about bias, what might they do?
Users are not usually in a position to evaluate the search engine’s
performance directly.71 Users making navigational queries can often tell
whether their intended destination is among the search results; users
making transactional queries, however, may be less able to tell whether the
engine has really directed them to the best sources. Some of this concern
may be alleviated by users’ ability to compare results from different search
engines, although other search engines may suffer from the same systematic
biases (think of the Chinese censorship and Googlearchy problems). Users
might respond by demanding additional information from a search engine
67. John Hiler, Google " Blogs—How Weblogs Influence a Billion Google Searches a Week,
MICROCONTENT NEWS, Feb. 26, 2002, http://www.microcontentnews.com/articles/googleblogs.
htm, available at http://web.archive.org/web/20030205011440/http://www.microcontentnews.
com/articles/googleblogs.htm; see also Ulken, supra note 64 (claiming inclusion of “non-
traditional news sources” in Google News creates increased prominence for extreme
viewpoints). Cf. Anil Dash, Nigritude Ultramarine, DASHES.COM, June 4, 2004,
http://www.dashes.com/anil/2004/06/04/nigritude_ultra (describing how collaboration
among blog authors to link Dash’s entry using keywords “nigritude ultramarine” nearly won an
SEO contest).
68. Matthew Hindman et al., “Googlearchy”: How a Few Heavily-Linked Sites Dominate
Politics on the Web 1 (July 28, 2003) (unpublished manuscript), available at
http://www.cs.princeton.edu/~kt/mpsa03.pdf.
69. Compare Junghoo Cho & Sourashis Roy, Impact of Search Engines on Page Popularity
1 (2004) (unpublished manuscript, available at http://oak.cs.ucla.edu/~cho/papers/cho-
bias.pdf) (“New and valuable pages are ignored just because they have not been given a chance
to be noticed by people.”), with Santo Fortunato et al., Topical Interests and the Mitigation of Search
Engine Bias, 103 PROC. NAT. ACAD. SCI. 12,684, Aug. 22, 2006, available at
http://cxnets.googlepages.com/PNAS_engine.pdf (“Yet, despite the rich-get-richer dynamics
implicit in the link analysis used to rank results, the use of search engines appears to mitigate
the average traffic attraction of high-degree pages.”).
70. See Clay Shirky, Power Laws, Weblogs, and Inequality, CLAY SHIRKY’S WRITINGS ABOUT THE
INTERNET (Feb. 8, 2003), http://www.shirky.com/writings/powerlaw_weblog.html (“Inequality
. . . is a reliable property that emerges from the normal functioning of the system.”). Notably,
even where scholars agree on the overall distribution of attention, they disagree on its
-1
implications. Compare Hindman et al., supra note 68 (x powerlaw distribution is a sign of lack of
-1
diversity), with CHRIS ANDERSON, THE LONG TAIL 126–27 (2006) (x powerlaw distribution is a
sign of diversity).
71. See, e.g, Diaz, supra note 3, at 147 (“The complexity and opacity of search technology
makes it almost impossible for users to notice what is ‘missing’ from their search results.”).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 23
about its ranking algorithms, this time to understand why it has made the
choices it has.72
Legally, search engines have a strong first line of defense against user
suits for bias in their browsewrap terms of service.73 Further, few business-
tort theories provide users with enforceable rights. One exception may be
the Federal Trade Commission’s (“FTC”) jurisdiction over misleading
business practices. Although it has not taken direct action against any search
engines, the FTC has communicated to search engines its belief that any
paid placement results should be clearly disclosed and distinguished from
organic search results.74 In cases of conscious, human-directed manipulation
of results, search engines may also have opened themselves to claims of
fraud on consumers by emphasizing the mechanical and supposedly
“objective” nature of their algorithms.75
Those who are concerned about systematic biases have also proposed
various forms of forced ranking or inclusions. One proposal would require
search engines to randomly intermix new content that has not yet had the
time to establish itself with older and already-popular content.76 Others
would require search engines to show users more diverse content to break
down their biases toward the familiar.77 There is a strong counterargument,
however, that regulators would be even more biased, as well as grossly
incompetent, at the task of dictating search results.78
72. But see James Grimmelmann, Note, Regulation by Software, 114 YALE L.J. 1719, 1734–38
(2005) (observing that some algorithms are so complicated that it may not be possible to say
“why” a computer made a particular decision).
73. See, e.g., Hubbert v. Dell Corp., 359 Ill. App. 3d 976, 983–84 (Ill. App. Ct. 2005)
(holding enforceable the arbitration clause in the “Terms and Conditions” on Dell’s web site).
74. See Letter from Heather Hippsley, Acting Associate Director, F.T.C. Division of
Advertising Practices, to [Search Engine Company] (June 27, 2002), available at
http://www.ftc.gov/os/closings/staff/commercialalertattatch.htm.
75. See Rebecca Tushnet, KinderStart: The Return, 43(B)LOG (Sept. 20, 2006), http://
tushnet.blogspot.com/2006/09/kinderstart-return.html (“If Google continues to tell searchers
one thing about how search results are generated and tell webmasters another, it might
behoove the FTC—the only entity with a realistic chance of affecting Google—to look into the
matter.”). But see Search King, Inc. v. Google Tech., Inc., No. CIV-02-1457-M, 2003 U.S. Dist.
LEXIS 27193, at *11 (D. Okla. May 27, 2003) (“Google and Page’s statements as to the
purported objectivity of the PageRank system cannot transform a subjective representation into
an objectively verifiable fact.”).
76. See Sandeep Pandey et al., Shuffling a Stacked Deck: The Case for Partially Randomized
Ranking of Search Engine Results, PROC. OF 31ST INT’L CONF. ON VERY LARGE DATABASES (VLDB)
(2005), available at http://oak.cs.ucla.edu/~cho/papers/cho-shuffle.pdf.
77. See SUNSTEIN, supra note 61, at 182–89 (arguing that “must carry” rules could be an
effective means of diversifying individual experiences); Susan L. Gerhart: Do Web Search Engines
Suppress Controversy?, FIRST MONDAY, Jan. 2004, http://firstmonday.org/issues/issue9_
1/gerhart/index.html (stating that a normal surfer would be “pulled” toward dominant sites
instead of more diverse ones).
78. See Eric Goldman, A Coasean Analysis of Marketing, 2006 WIS. L. REV. 1151.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
24 93 IOWA LAW REVIEW [2007]
B. PROVIDERS’ INTERESTS
1. Minimizing Costs
Turning to providers, we start with the first step in search: indexing.
Every time a search engine asks to index content from a provider, the
provider must use a little server time and network bandwidth to respond.
Every time a user requests content from the provider, the provider must
similarly expend a little server time and bandwidth. Multiply by a large
collection of content, many search engines, frequent indexing, and many
users—and many a mickle makes a muckle.79
Most of the time, providers willingly cooperate: getting lots of users
from search engines is usually a good thing, well worth the technical costs.
Still, some providers object. Their reasons vary. Providers may be bothered
by inefficient search engines that index too often or with unwelcome
burstiness. They may want to be available online to users but not to be
searchable.80 They may find that flows of user traffic from search engines
don’t include the sorts of users for whom they’re looking.81 Some of them
may simply welcome any leverage they can employ to force search engines to
pay for the privilege of indexing.82
Turning to the law, providers have attempted to defend their servers
from unnecessary indexing burdens on three principal grounds. First, they
have used the common-law tort of trespass to chattels. Despite some
successes for this theory,83 the 2003 California Supreme Court’s heavily
publicized holding in Intel v. Hamidi 84 (that the tort would not lie without a
showing of harm to the chattel or the owner’s ability to use it) would largely
79. In 2006, forty-six percent of all requests for pages from this author’s web site came
from the Yahoo! robot.
80. See generally danah boyd, Facebook’s “Privacy Trainwreck”: Exposure, Invasion, and Drama,
Sept. 8, 2006, http://www.danah.org/papers/FacebookAndPrivacy.html (arguing that
individuals may be more comfortable revealing personal information when they do not expect
it to be easily searchable); see also infra Part II.B.2 (discussing unfair-competition reasons why
providers may not like searchability).
81. In the so-called “Slashdot effect,” being recommended by a popular web site produces
a sudden and huge influx of traffic. Not only does the linked site face a potentially huge
bandwidth bill, but it also runs the risk of seeing its servers crash under the load, meaning that
it reaches none of the users it is trying to. For more on the Slashdot effect, see generally Jason
Kottke, Digg vs. Slashdot (or, Traffic vs. Influence), KOTTKE.ORG, Jan. 12, 2006,
http://www.kottke.org/06/01/digg-vs-slashdot.
82. See Danny Sullivan, Google’s Belgium Fight: Show Me the Money, Not the Opt-Out, Say
Publishers, SEARCH ENGINE WATCH (Sept. 20, 2006), http://blog.searchenginewatch.
com/blog/060920-152314 (analyzing the motivations of publishers suing Google to prevent
indexing).
83. See generally Register.com, Inc. v. Verio, Inc., 356 F.3d 393 (2d Cir. 2004); Oyster
Software, Inc. v. Forms Processing, Inc., No. C-00-0724, 2001 WL 1736382 (N.D. Cal. Dec. 6,
2001); eBay, Inc. v. Bidder’s Edge, Inc., 100 F. Supp. 2d 1058 (N.D. Cal. 2000).
84. Intel Corp. v. Hamidi, 71 P.3d 296, 308–09 (Cal. 2003).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 25
preclude a trespass-to-chattels claim, unless the search engine was
egregiously burdensome.85
Providers have had more luck under sui generis state and federal
computer-intrusion statutes, which generally prohibit access to a computer
system without authorization. Here, courts have been willing to say that any
use of a server—including spidering—that the owner does not condone is
ipso facto unauthorized. Despite academic criticism86 and at least one court’s
discomfort with the theory,87 courts considering claims against search
engines under the federal Computer Fraud and Abuse Act’s civil provisions88
have held for the providers.89
Third, providers have at times alleged the existence of a contract
prohibiting a search engine from spidering their content. In one notable
2004 case, the Second Circuit found that displaying a notice on information
returned by a server forbidding certain uses of that information was
sufficient to bind even a company accessing the information purely through
a spidering program.90 In reasoning directly applicable to search engines,
the court held that the repeated access meant that knowledge of the
purported contractual terms should be imputed to the spider’s operator.91
The leading case on consumer interactions with contracts presented on the
web, Specht v. Netscape Communications Corp., is in accord. The consumer in
that case was not bound by the contract but only because the contract was
not clearly displayed on the web page in a way that forced the user to see it
or acknowledge it before clicking on a download link. The court left little
doubt that a provider with sufficient willingness could craft terms and an
interface for displaying them that would bind users who clicked through.92
None of these theories has been settled definitively. Some of the
analyses are cursory at best, and none has found liability for a general-
purpose search engine, rather than a specialized service that could be
characterized as a direct competitor to the aggrieved provider. Nonetheless,
85. See also Ticketmaster Corp. v. Tickets.com, Inc., No. 99-7654, 2003 WL 21406289, at *3
(C.D. Cal. Mar. 7, 2003) (requiring a showing of actual harm “pending appellate guidance”).
86. See generally Orin Kerr, Cybercrime’s Scope: Interpreting “Access” and “Authorization” in
Computer Misuse Statutes, 78 N.Y.U. L. REV. 1596 (2003).
87. See Lockheed Martin Corp. v. Speed, No. 6:05.cv-1580-Orl-31KRS, 2006 U.S. Dist.
LEXIS 53108, at *18–19 (M.D. Fla. Aug. 1, 2006).
88. 18 U.S.C. § 1030(g) (2000).
89. See EF Cultural Travel BV v. Explorica, Inc., 274 F.3d 577, 585 (1st Cir. 2001); SW.
Airlines Co. v. Farechase, Inc., 318 F. Supp. 2d 435, 438–41 (N.D. Tex. 2004).
90. See Register.com, Inc. v. Verio, Inc., 356 F.3d 393, 393 (2d Cir. 2004).
91. In one remarkable case, a court held that a plaintiff’s allegations of a binding
browsewrap contract could survive a motion to dismiss, even when the contract’s terms were not
displayed in a manner that would signal to a spider’s creator that the provider was attempting to
form a contract. See Internet Archive v. Shell, No. 06-cv-01726-LTB-CBS, 2007 U.S. Dist. LEXIS
10239, at *18–19 (D. Colo. Feb. 13, 2007).
92. Specht v. Netscape Comm. Corp., 306 F.3d 17, 35 (2d Cir. 2002).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
26 93 IOWA LAW REVIEW [2007]
the trend seems to be that a search engine could be prohibited from
indexing an unwilling provider, at least on contractual grounds.
On the web, the matter appears to have settled into a rough
equilibrium, with most providers using robot exclusion protocols to inform
the public which robots are allowed to spider which portions of their
content. Most major search engines respect these requests.93 This informal
compromise does not directly bind users, however, because the norms of
robot exclusion protocols apply only to the operators of indexing robots.94
So much for indexing. Compare the legal regime governing excessive
user attention. It is well-established that deliberately orchestrated denial-of-
service attacks are both crimes and torts,95 and normal principles of co-
conspirator liability suggest that recruiting thousands of others to act in
concert to attack a computer system might make the recruiter (and quite
possibly each user) liable for all resulting harm.96 The more interesting cases
involve intermediaries (such as search engines) that overwhelm a site by
directing to it users who don’t know they’re part of an online mob.
At least one case has held that a person who deliberately manipulates
the inputs to an information-location tool so that a provider is incorrectly
listed may be liable in trespass to chattels.97 Similarly, a search engine that
misleads users as to what they will find may be indirectly intermeddling and,
thus, also liable in trespass to chattels.98 A substantial gray area remains. On
the one hand, any result returned by a search engine involves an intent to
cause the user to access the site (and thus to use the chattel of the server);
on the other, even an intention by a search engine to cause harm to a site
may not be accompanied by any deception of users.99
93. See Patricia Bellia, Defending Cyberproperty, 79 N.Y.U. L. REV. 2164, 2250–51 (2004)
(arguing that this result is, for the most part, efficient and just).
94. See Eric J. Feigin, Architectures of Consent: Internet Protocols and Their Legal Implications, 56
STAN. L. REV. 901, 916–17 (2004) (arguing that Internet norms of access and restriction
embedded in technical protocols are entitled to legal respect).
95. See Computer Fraud and Abuse Act, 18 U.S.C. § 1030(c)(2)(B)(ii) (2000); United
States v. Ancheta, No. CR 05-1060 (C.D. Cal. Feb. 2005) (indictment), available at
http://fl1.findlaw.com/news.findlaw.com/hdocs/docs/cyberlaw/usanchetaind.pdf.
96. See U.S. SENTENCING GUIDELINES MANUAL § 1B1.3(a)(1)(B) (2006) (stating that the
base offense level for jointly undertaken activity is based on “all reasonably foreseeable acts and
omissions of others in furtherance of the jointly undertaken criminal activity”).
97. Sch. of Visual Arts v. Kuprewicz, 771 N.Y.S.2d 804, 808 (N.Y. Sup. Ct. 2003) (holding
that one who subscribed another to high-volume e-mail lists could be liable in trespass to
chattels).
98. See RESTATEMENT (SECOND) OF TORTS § 217 cmt. e (1965) (“[A] trespass may be
committed by causing a third person through duress or fraud to intermeddle with another’s
chattel.” (emphasis added)).
99. On the difficulties involved in assessing intent and effects, compare Universal Tube &
Rollform Equip. Corp. v. YouTube, Inc., No. 3:06CV02628, 2007 U.S. Dist. LEXIS 40395, at
*19–20 (N.D. Ohio June 4, 2007) (holding that choosing the “youtube.com” domain name does
not involve “duress or fraud” in causing users to visit “utube.com”), with Dave Plonka, Flawed
Routers Flood University of Wisconsin Internet Time Server (Aug. 21, 2003), http://www.cs.wisc.edu/
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 27
One final cost requires discussion: Money spent on search engine
advertising could have been spent on some other useful cause. Search
advertising campaigns are increasingly managed like any other advertising
campaign, often as part of a comprehensive marketing strategy—and with
the same advertiser pressure for lower rates.100 Since users could also be
directed to a provider through organic search results, search advertising is a
partial substitute for search rankings. There have been occasional
accusations that search engines churn their rankings or deliberately demote
some providers to spur them to purchase search advertising.101
2. Avoiding Unfair Competition
Search engines can help users to acquire content in new ways. Some of
those ways disrupt providers’ preferred relationships with users. Does the
fact that the content originally came from providers give them a right to
demand that search engines only present it to users in ways that providers
approve of? Answering such a question is the domain of unfair-competition
law, which defines the limits of businesses’ rights to prevent others from
diverting customers. Search engines raise questions that blur the boundaries
between copyright, trademark, and common-law theories of
misappropriation.102 This Section analyzes unfair-competition arguments as
they apply to the different information flows in search, but the area is
sufficiently unstable that this taxonomy is a matter of convenience rather
than settled law.
Start with indexing. A search engine’s spidering processes require
making at least one initial copy of any content the engine wishes to index;
providers have complained that this initial copy is unauthorized and hence
~plonka/netgear-sntp/ (discussing a company that shipped thousands of home networking
devices that accessed a university server for its intended purpose but in far greater numbers
than appropriate).
100. See generally Search Engine Marketing Professional Organization (“SEMPO”), About
SEMPO, http://www.sempo.org/about.
101. Evaluating such charges is difficult, given the secrecy surrounding search ranking
algorithms. Independent search analyst Danny Sullivan argues plausibly that the churn in
rankings is a side effect of anti-SEO efforts, rather than a deliberate strategy. See Danny Sullivan,
What Happened to My Site on Google?, SEARCH ENGINE WATCH, Dec. 7, 2003,
http://searchenginewatch.com/showPage.html?page=3286101 (“[T]here are far easier ways
that Google could boost ad revenue uptake without doing sneaky, behind-the-scene actions
. . . .”).
102. The difference between the unfair-competition concerns discussed in this Section,
Part II.B.2, and the unwanted-access concerns discussed above, supra Part II.B.1, is that unfair
competition deals with a loss of human attention from users, while unwanted access has to do
with technical burdens. The difference between unfair-competition concerns and the placement
concerns discussed below, infra Part II.B.3, is that the unfair-competition concerns are about
competition from search engines, while the placement concerns are about competition with other
providers. In the former, the search engine itself steals users; in the latter, it shows undue
favoritism among providers.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
28 93 IOWA LAW REVIEW [2007]
infringes the provider’s copyright. This is the principal doctrinal peg on
which the Authors’ Guild hangs its copyright suit against the Google Book
indexing project.103 Because indexing by itself doesn’t involve users and thus
harms only an author’s abstract interest in controlling her work, courts have
tended to find it fair if they approve of the subsequent purposes to which
the engine puts the copies—and unfair if they don’t.104 Providers, however,
have begun to argue that even copies retained but not shown to users are
troublesome because of the risk of a security breach exposing the archives to
bulk copying.105
Next, consider the case in which the search engine directly provides
users with content. We could describe this provision in copyright terms; the
search engine is arguably making and publicly distributing unauthorized
copies.106 Unauthorized caching suggests infringement; providers who sell
the content or show it with advertising see their business models disrupted
by a search engine that serves up the content.107 Major search engines
generally honor requests not to cache, but they have forced providers to use
standard technical measures to make those requests.108 Giving users content
103. See Complaint at 2, Authors’ Guild v. Google, Inc., No. 05-CV-8136 (S.D.N.Y. Sept. 20,
2005), available at http://pub.bna.com/eclr/05cv8136comp.pdf (calling indexing “massive
copyright infringement”).
104. See Kelly v. Arriba Soft Corp., 336 F.3d 811, 822 (9th Cir. 2003) (finding fair use). But
see UMG Recordings, Inc. v. MP3.com, Inc., 92 F. Supp. 2d 349, 352–53 (S.D.N.Y. 2000) (finding
no fair use). In these cases, the line between providers and third parties is at its least distinct.
See, e.g., Sony Computer Entm’t, Inc. v. Connectix Corp., 203 F.3d 596, 608 (9th Cir. 2002)
(finding fair use in initial copying for purposes of reverse engineering).
105. See Fair Use: Its Effects on Consumers and Industry: Hearing Before the H. Subcomm. on
Commerce, Trade, and Consumer Protection of the H. Comm. on Energy and Commerce, 109th Cong. 66
(2005) (testimony of Paul Aiken, Executive Director, Authors Guild) (“Since there’s no license
needed, in Google’s view, Google doesn’t have to give rightsholders contractual assurances of
the security of their database.”).
106. See 17 U.S.C. § 106(1), (3) (2000) (providing certain exclusive rights to copyright
owners, including reproduction and distribution).
107. Cf. Twentieth Century Fox Film Corp. v. iCraveTV, No. 00-121, 2000 U.S. Dist. LEXIS
1013, at *2 (W.D. Pa. Jan. 28, 2000) (granting an injunction against Internet sites framing TV
stations’ live feeds with advertisements and rebroadcasting them on the Internet).
108. See Field v. Google, Inc., 412 F. Supp. 2d 1106, 1122–23 (D. Nev. 2006); Internet
Archive, Removing Documents from the Wayback Machine, http://www.archive.org/
about/exclude.php. But see Rebecca Bolin, Locking Down the Library, 29 HASTINGS COMM. & ENT.
L.J. 1, 9–19 (2006) (arguing that copyright-related removal from online archives threatens
preservation of human memory). Providers and search engines have both been developing
technical measures to improve the precision of signals in this area. See Jeffrey Goldfarb,
Publishers Aim for Some Control of Search Results, INT’L BUS. TIMES, Sept. 22, 2006,
http://ibtimes.com/articles/20060922/global-publishers-google-search-engines-copyright.htm
(detailing an attempt by publishers to develop an automated system to grant content-use
permission); The Web Robots Pages, The Web Robots FAQ, http://
www.robotstxt.org/wc/faq.html. But see Niva Elkin-Koren, What Contracts Cannot Do: The Limits of
Private Ordering in Facilitating a Creative Commons, 74 FORDHAM L. REV. 375, 407–17 (2005)
(arguing that increased precision of permissions increases the perception that permission is
required for any reuse).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 29
directly may also interfere with proper attribution by severing the link
between source and content. This objection appeals to trademark policies,109
but trademark causes of action, ironically, may be hemmed in by
copyright.110
The issue becomes murkier in the case of thumbnailing and other
practices that give users only excerpts or summaries of content. Some
authority holds that the thumbnails are protected fair use;111 this holding is
not clearly established, however, and drawing the line between thumbnail
and full copy (or derivative work) may require significant case-by-case
analysis.112 Some providers, moreover, have argued that thumbnailing—or
even offering search itself—involves the exploitation of value produced by
content and properly attributable to that content.113 Precisely because this
related market is valuable, goes the argument, the content owners should
have the exclusive right to exploit it.114
Even when users obtain content directly from providers, the search
engine may change how they obtain it, through techniques such as framing,
deep linking, and inlining. If the engine provides the user with specific
technical instructions for obtaining content, the user may experience the
content in a context not intended by providers, even though the provider
itself supplied the content to the user upon request. Providers have been
particularly upset when search engines cause users to bypass advertising on
the provider’s site.115 United States courts have not been able to agree either
109. See 15 U.S.C. § 1125(a)(1) & (a)(1)(A) (2000) (making actionable a “false designation
of origin . . . likely to cause confusion”).
110. See generally Dastar Corp. v. Twentieth Century Fox Film Corp., 539 U.S. 23 (2003)
(holding that authorship is not “origin” for purposes of Section 1125(a)). But see F. Gregory
Lastowka, The Trademark Function of Authorship, 85 B.U. L. REV. 1171, 1172 (2005) (“[T]he
Dastar approach is misguided.”).
111. See Kelly v. Arriba Soft Corp., 336 F.3d 811, 822 (9th Cir. 2003).
112. See Perfect 10 v. Amazon.com, 487 F.3d 701 (9th Cir. 2007) (holding thumbnailing
likely to be a fair use as against a third-party copyright holder).
113. See, e.g., Nick Taylor, . . . But Not at Writers’ Expense, WASH. POST, Oct. 21, 2005, at A18
(“The alphabet ought to be free, most certainly, but the people who painstakingly arrange it
into books deserve to be paid for their work.”). But see Emily Anne Proskine, Note, Google’s
Technicolor Dreamcoat: A Copyright Analysis of the Google Book Search Library Project, 21 BERKELEY
TECH. L.J. 213, 231–32 (2006) (“None of copyright’s exclusive rights suggest that publishers or
authors should possess a monopoly over the indexing and searching of their works.”).
114. But see Rochelle Dreyfus, Expressive Genericity: Trademarks as Language in the Pepsi
Generation, 65 NOTRE DAME L. REV. 397, 405 (1990) (“Furthermore, fallacies in the fundamental
assumptions made by courts that have approved this ‘if value, then right’ theory mean that the
right lacks a coherent limit.”).
115. See, e.g., Complaint at ¶36, Wash. Post Co. v. TotalNews, Inc., No. 97 Civ. 1190
(S.D.N.Y. Feb. 20, 1997), available at http://legal.web.aol.com/decisions/dlip/washcomp.html
(“Yet an advertisement on one of Plaintiffs’ sites, when seen through the totalnews.com window,
is reduced in size, may even be totally obscured by the totalnews frame, and is forced to
compete for the user’s attention with the visual clutter of the totalnews.com frame, including
other advertising -- possibly including advertising for directly competitive products.”).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
30 93 IOWA LAW REVIEW [2007]
on the copyright116 or trademark117 implications of these techniques, but
European courts have been willing to find that deep linking to news violates
the European Database Directive.118 An additional set of complications arises
when the relevant content is uncopyrightable (e.g., because it consists of
unprotectable fact, rather than protectable expression). Here, providers may
wish to complain both of direct provision and altered presentation, but they
must steer clear of copyright preemption.119
Finally, these questions cannot be addressed without considering users’
rights. Not only do users enjoy significant statutory and fair-use rights to
make copies for their own caching and archival purposes,120 but there are
strong arguments that many of the transformations to which search engines
subject content would be wholly legal if carried out by users directly. The
role of the search engine as an intermediary carrying out those activities for
them raises deeply contentious issues of intellectual-property policy.121
116. See Kelly v. Arriba Soft Corp., 280 F.3d 934, 944–47 (9th Cir. 2002), withdrawn by, 336
F.3d 811 (9th Cir. 2003) (finding that deep linking and framing infringe on the exclusive
public-display right).
117. See Digital Equip. Corp. v. AltaVista Tech., 960 F. Supp. 456, 478–79 (D. Mass. 1997)
(enjoining the defendant from linking to the plaintiff and creating a false impression of
affiliation); Knight-McConnell v. Cummins, No. 03 Civ. 5035 (NRB), 2004 U.S. Dist. LEXIS
14746, at *7 (S.D.N.Y. July 29, 2004) (“The mere appearance on a website of a hyperlink to
another site will not lead a web-user to conclude that the owner of the site he is visiting is
associated with the owner of the linked site.”).
118. See Copiepresse v. Google, Inc., Tribunal de Premiere Instances de Bruxelles [Court
of First Instance] Brussels, 13 février 2007, J.B.C. 7964 (Belg.) (prohibiting deep linking),
available at http://www.copiepresse.be/13-02-07-jugement-en.pdf; Press Release,
Newsbooster.com, Newsbooster Keeps on Fighting (July 12, 2002), available at
http://www.chillingeffects.org/news.cgi?NewsID=199 (discussing Danish Newspaper
Publishers’ Association v. Newsbooster.com (2002) (Denmark), which prohibited
Newsbooster.com’s deep linking). But see Zoekallehuizen.nl/NVM, Rechtbank Arnhem [Court
Arnhem], 16 maart 2006, LJN AV5236 (Neth.) (allowing deep linking), available at
http://zoeken.rechtspraak.nl/resultpage.aspx?snelzoeken=true&searchtype=ljn&ljn= AV5236&
u_ljn=AV5236. For a discussion of international cases relating to deep linking, see Linking
Cases—Deep Links/Search Engines, http://www.linksandlaw.com/linkingcases-deeplinks-
3.htm (summarizing and linking to Home v. Ofir (2006) (Denmark) (allowing deep linking),
Zoekallehuizen.nl v. NVM (2006) (Neth.) (same), and Bixee.com v. Nahuri.com (2006) (Ind.)
(prohibiting deep linking)).
119. See Feist Publ’ns, Inc. v. Rural Tel. Serv. Co., Inc., 499 U.S. 340, 364 (1991) (holding
that a phone book was not original enough to be protected by copyright); Ticketmaster Corp. v.
Tickets.com, Inc., No. CV 99-7654 HLH(BQRX), 2000 WL 525390, at *2, *4 (C.D. Cal. March
27, 2000) (applying Feist).
120. See Sony Corp. of Am. v. Universal City Studios, Inc., 464 U.S. 417, 454–55 (1984)
(holding that a VCR owner’s taping of television shows is fair use). See generally Jessica Litman,
Lawful Personal Use, 85 TEX. L. REV. 1871 (2007) (discussing personal uses of copyrighted
material, such as private home copying).
121. See, e.g., Marvel Enters. v. NCSoft Corp., No CV 04-9253-RGK (PLAx), 2005 U.S. Dist.
LEXIS 8448, at *11–17 (C.D. Cal. March 9, 2005) (considering copyright and trademark
implications of actions by users that would have been lawful private uses offline); Paramount v.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 31
3. Prominent Placement in Results
Prominent search-result placements carry immense value. Users are
more likely to click on the first result than the second, the second than the
third, and so on. If you don’t appear on the first few pages of results, you
may as well not exist. Accordingly, providers seek the highest possible
placement. Some turn to SEO. Others turn to the law.
When a search engine returns result R in response to query Q, it
effectively asserts that “R is a good source for information about Q,” or
perhaps that “R is a good place to acquire Q.” Such assertions implicate the
policies of trademark and advertising law, which seek to keep consumers
from being confused by misleading claims about the relationship between
businesses and the goods they offer.
Trademark law has taken a strong interest in Internet information-
location tools. Trademark holders have been able to appeal to ordinary
principles of trademark law, as well as to various sui generis legal regimes
dealing with domain names,122 against those who use confusing domain
names to draw away consumers.123 Courts have split on whether purchasing
search advertisements tied to queries on a competitor’s trademark can
constitute infringement.124 In a decision that has attracted substantial
criticism,125 the Ninth Circuit held that the use of trademarks in hidden
metatags (an early, albeit probably ineffective, form of SEO) could be
actionable as causing “initial interest confusion.”126
Three years later, the Ninth Circuit extended the doctrine so that
liability could run against search engines themselves.127 The court reasoned
ReplayTV, 298 F. Supp. 2d 921, 923 (C.D. Cal. 2004) (objecting to a feature of a personal video
recorder allowing consumers to skip commercials automatically).
122. See Internet Corporation for Assigned Names and Numbers, Uniform Domain Name
Dispute Resolution Policy (“UDRP”) (1999), http://www.icann.org/udrp/udrp-policy-
24oct99.htm; see also Anti-Cybersquatting Consumer Protection Act (“ACPA”), 15 U.S.C. §
1125(d) (2000); id. § 1117. See generally Lamparello v. Falwell, 420 F.3d 309 (4th Cir. 2005).
123. See generally, e.g., People for the Ethical Treatment of Animals v. Doughney, 263 F.3d
359 (4th Cir. 2001). But see Lamparello, 420 F.3d at 316#17 (distinguishing Doughney).
124. Compare Edina Realty, Inc. v. TheMLSonline.com, No. Civ. 04-4371, 2006 WL 737064,
at *7 (D. Minn. Mar. 20, 2006) (yes), and Buying for the Home, LLC v. Humble Abode, LLC,
459 F. Supp. 2d 310, 332 (D.N.J. 2006) (yes), and J.G. Wentworth, S.S.C. v. Settlement Funding
LLC, No. 06-0597, 2007 U.S. Dist. LEXIS 288, at *8 (E.D. Pa. Jan. 4, 2007) (yes), with Site Pro-1
v. Better Metal, LLC, No. 06-6508, 2007 U.S. Dist. LEXIS 34107, at *5#6 (E.D.N.Y. May 9, 2007)
(no), and Merck & Co., Inc. v. Mediplan Health Consulting, Inc., 425 F. Supp. 2d 402, 415#16
(S.D.N.Y. 2006) (no), and Rescuecom Corp. v. Google, Inc., 456 F. Supp. 2d 393, 403#04
(N.D.N.Y. 2006) (no).
125. See Eric Goldman, Deregulating Relevancy in Internet Trademark Law, 54 EMORY L.J. 507,
565 (2005); see also Playboy Enters., Inc. v. Netscape Commc’ns Corp., 354 F.3d 1020, 1034#36
(9th Cir. 2004) (Berzon, J., concurring).
126. Brookfield Commc’ns, Inc., v. W. Coast Entm’t Corp., 174 F.3d 1036, 1066 (9th Cir.
1999).
127. Playboy, 354 F.3d at 1024#29.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
32 93 IOWA LAW REVIEW [2007]
that a search engine’s use of a trademarked term as an advertising keyword
was a use of the trademark capable of causing confusion, whether or not the
keywords were used in the advertisements themselves.128 By way of contrast,
other courts have held that adware vendors, who sell software that displays
advertisements based on the web pages visited by users, do not violate the
trademark rights of the providers of the underlying web pages.129
These precedents provide conflicting guidance on the obligations of
search engines that trigger advertisements based on search queries
containing trademarked terms, and search engines have vacillated in their
policies on selling such advertisements.130 Google, which has been the most
vigorous of the major search engines in contesting suits by trademark
holders, has both won and lost motions to dismiss on nearly identical facts.131
Utah has amended its trademark law to prohibit many keyword-triggered
advertisements,132 although the prohibition raises substantial constitutional
and preemption issues.133 So far, no search engine has been sued for
returning a competitor’s web page as an organic search result in a search on
a trademarked term, but such suits cannot be ruled out. Some
128. Id. at 1025–26.
129. 1-800-Contacts, Inc. v. WhenU.com, Inc., 414 F.3d 400, 410#12 (2d Cir. 2005); Wells
Fargo & Co. v. WhenU.com, Inc., 293 F. Supp. 2d 734, 772 (E.D. Mich. 2003); U-Haul Int’l, Inc.
v. WhenU.com, Inc., 279 F. Supp. 2d 723, 728#29 (E.D. Va. 2003). These cases also raised
copyright theories tied to the allegedly altered display of the web pages.
130. Compare Google, AdWorlds Trademark Complaint Procedure, http://www.
google.co.uk/tm_complaint_adwords.html (forbidding only the use of the trademark in ad
text), with Yahoo! Search Marketing, Trademarks, http://searchmarketing.yahoo.com/
legal/trademarks.php (forbidding bidding on trademarked terms except under stated
conditions).
131. See generally Google, Inc. v. Am. Blind & Wallpaper Factory, Inc., No. C 03-5340 JF
CRSJ, 2007 U.S. Dist. LEXIS 32450 (N.D. Cal. Apr. 18, 2007) (allowing a theory of infringement
for use in keywords to proceed on motion for summary judgment); Gov’t Employees Ins. Co. v.
Google, Inc., No. 01:04cv507, 2005 U.S. Dist. LEXIS 18642 (E.D. Va. Aug. 8, 2005) (finding for
the search engine in a bench trial); Google, Inc. v. Am. Blind & Wallpaper Factory, Inc., No. 03-
05340, 2005 U.S. Dist. LEXIS 6228 (N.D. Cal. Mar. 30, 2005) (allowing a theory of infringement
for use in keywords to proceed on motion to dismiss); Gov’t Employees Ins. Co. v. Google, Inc.,
330 F. Supp. 2d 700 (E.D. Va. 2004) (dismissing a theory of trademark infringement for use in
advertising keywords but allowing a theory of infringement for use in advertising text to
proceed). See also 800-JR-Cigar, Inc. v. GoTo.com, 437 F. Supp. 2d 273, 297 (D.N.J. 2006)
(allowing a keyword-use theory to proceed); Elinor Mills, Google Loses French Trademark Lawsuit,
CNET NEWS, June 28, 2006, http://www.news.com/2100-1030_3-6089307.html (reporting on
similar French decisions in favor of trademark holders).
132. Trademark Protection Act, § 70-3a-402(1)(c), S.B. 236, 2007 Gen. Sess. (Utah 2007),
available at http://le.utah.gov/~2007/bills/sbillamd/sb0236.htm.
133. See Eric Goldman, Utah Bans Keyword Advertising, TECH. & MKTG. L. BLOG (Apr. 3,
2007), http://blog.ericgoldman.org/archives/2007/04/utah_bans_keywo.htm (stating that the
law raises Dormant Commerce Clause and First Amendment concerns).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 33
commentators have proposed that a similar result be required by
regulation.134
Providers aggrieved by their poor placement have also sued search
engines for business libel and related theories.135 The providers generally
have argued either that (1) their content was highly relevant by any objective
standard and that a poor ranking is, in effect, a lie, or (2) the search engine
reduced the providers’ rankings out of malice.136 These state-law claims have
not fared well in court. It may not have helped some providers’ legal cases
that their prelitigation actions looked a lot like black-hat SEO.137
C. THIRD PARTIES’ INTERESTS
We turn now to third parties’ interests in suppressing certain content
flows. Bringing these third parties into the picture complicates the politics.
The flow of content from providers to search engines to users will simply
stop if any of those three groups pulls out entirely. They may argue over the
division of spoils and jockey for relative advantage, but they are united in
wanting to use search to make information more usefully accessible. Third
parties, on the other hand, object to the content flow itself. This creates an
adversarial dynamic: third party versus everyone else—with some interesting
fault lines within the “everyone else” coalition.
1. Ownership
We begin with intellectual-property interests. Here, providers and users
have a common interest in the flow of infringing content. Search engines
have a business interest in serving them, counterbalanced by an interest in
avoiding implication in the infringement. Note that there is some doctrinal
134. See, e.g., Pasquale, supra note 3, at 136 (proposing that trademark holders be allowed
to place an asterisk next to unsanctioned search results); James A. Rossi, Protection for Trademark
Owners: The Ultimate System of Regulating Search Engine Results, 42 SANTA CLARA L. REV. 295, 347–
54 (proposing that search engines should always offer the option of returning results that the
holder of the trademarked search term selects).
135. See generally KinderStart.com, LLC v. Google, Inc., No. C 06-2057 JF CRSJ, 2007 U.S.
Dist. LEXIS 22637, at *3 (N.D. Cal. Mar. 16, 2007); Langdon v. Google, Inc., 474 F. Supp. 2d
622, 626 (D. Del. 2006); Roberts v. Google, Inc., No. 1-06-CV-063047 (Cal. Sup. Ct. May 12,
2006) (voluntary dismissal); Search King, Inc., v. Google Tech., Inc., No. CIV-02-1457-M, 2003
U.S. Dist. LEXIS 27193, at *3 (W.D. Okla. May 27, 2003); Complaint, Datner v. Yahoo! Inc., No.
BC355217 (Cal. Sup. Ct. July 11, 2006); Complaint, CLRB Hanson Industries, LLC v. Google,
Inc., No. 1-05-CV-046409 (Cal. Sup. Ct. Aug. 3, 2005).
136. Search King, Inc., 2003 U.S. Dist. LEXIS 27193, at *8–9.
137. See Dahlia Lithwick, Google-Opoly: The Game No One but Google Can Play, SLATE, Jan. 29,
2003, http://www.slate.com/id/2077875/ (“SearchKing in effect has its clients collude to trick
Google into boosting everyone’s ratings.”). Cf. David Kesmodel, Blogger Faces Lawsuit Over
Comments Posted by Readers, WALL ST. J. ONLINE, Aug. 31, 2005, http://online.
wsj.com/pubic/article/SB112541909221726743-_vX2YpePQV7AOIl2Jeebz4FAfS4.20060831.
html?mod=blogs (describing a defamation lawsuit against a blogger whose commenters
criticized the plaintiff’s SEO tools).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
34 93 IOWA LAW REVIEW [2007]
overlap between these claims by third parties and the unfair-competition
claims providers might bring against search engines.138
Starting, as above, with indexing, the initial copies a search engine
makes while spidering seem to be unambiguous fair uses.139 So far, search
engines have also been doing well at the other end of the chain: courts have
held that neither linking to infringing content nor framing it constitutes
direct infringement and that thumbnailing is a fair use.140 Caching and
archiving are riskier; most major search engines behave cautiously, removing
allegedly infringing content from their caches.141
Secondary copyright liability is murkier. Grokster teaches that the makers
of infringement-facilitating technologies must both pass the Sony staple-
article-of-commerce test and steer clear of purposeful, culpable inducement
of infringement.142 Given that Grokster and its brethren consisted of a
search application fused with a file-transfer application, this holding applies
directly to search engines.143 The principal web search engines easily pass
the “capable of substantial noninfringing uses” prong of the Sony test, but
more specialized search engines may not. As for purposeful, culpable
inducement, the application of this language to many technologies, not just
search engines, remains unclear.144
The immunities and subpoena processes detailed in Section 512 of the
Digital Millennium Copyright Act (“DMCA”)145 also will be significant in
search engine copyright litigation. ISP litigation has clarified some of the
issues, but the courts have not yet extensively glossed Section 512(d), the
immunity for “Information Location Tools.” That section incorporates by
138. See supra Part II.B.2.
139. See Kelly v. Arriba Soft Corp., 280 F.3d 934 (9th Cir. 2002), withdrawn by 336 F.3d 811,
822 (9th Cir. 2003) (finding that reproduction of thumbnail images is a fair use).
140. See Perfect 10, Inc. v. Amazon.com, Inc., 487 F.3d 701 (9th Cir. 2007).
141. See, e.g., MSN.com, Site Owner Help: Control Which of Your Pages are Indexed,
http://search.msn.com.sg/docs/siteowner.aspx?t=SEARCH_WEBMASTER_REF_RestrictAccess
ToSite.htm. (“Prevent MSNBot from caching a page”). But see 17 U.S.C. § 512(b) (2000)
(providing a safe harbor for “system caching”); Parker v. Google, Inc., 422 F. Supp. 2d 492 (E.D.
Pa. 2006), aff’d, No. 06-3074, 2007 WL 1989660, at *3 (3d Cir. July 10, 2007) (per curiam)
(holding that archiving is not “volitional” and, therefore, is not copyright infringement).
142. MGM Studios, Inc. v. Grokster, Ltd., 545 U.S. 913, 935 (2005). Sony held that one who
manufactures and distributes a technology will not be liable for infringements committed by its
users as long as the technology has “substantial noninfringing uses.” Sony Corp. v. Universal
City Studios, Inc., 464 U.S. 417, 456 (1984). Grokster added to that test a requirement that the
maker must not have acted with the intent of inducing its users to infringe. Grokster, 545 U.S. at
935.
143. See Grokster, 545 U.S. at 921 (describing Grokster’s search function).
144. Compare id. at 919 (“[O]ne who distributes a device with the object of promoting its
use to infringe copyright, as shown by clear expression or other affirmative steps taken to foster
infringement, is liable” (emphasis added)), with MGM Studios, Inc. v. Grokster, Ltd., 454 F. Supp.
2d 966, 985 (C.D. Cal. 2006) (“Plaintiffs need prove only that StreamCast distributed the
product with the intent to encourage infringement.”).
145. Digital Millennium Copyright Act, 17 U.S.C. § 512 (2000).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 35
reference the ISP-focused notice-and-takedown procedure for “Information
Residing on Systems or Networks at Direction of Users” described in Section
512(c), and search engines have been diligent about removing links for
which they receive notices alleging copyright infringement.146 But the
parallel is not exact. On the one hand, because search engines do not have
the direct relationship with users that hosting services have, providers are
more vulnerable to abuse of the notice-and-takedown process. On the other
hand, because the search engine’s role in such cases is generally only to link
to information, and because the notices must specify the “reference or link,
to material or activity claimed to be infringing,” search engines can generally
undermine a takedown notice by displaying the notice itself.147 It also
remains unclear how other provisions of Section 512, such as the
termination-of-repeat-infringers requirement of Section 512(i) and the
subpoena-to-identify-infringers provision of Section 512(h), apply to search
engines.148
Cutting across all of these copyright issues are the general problems of
what it means for a search engine to have “knowledge” of infringement and
the extent to which a search engine profits from infringing activity or can
control infringers.149 Napster was charged with knowledge of particular
infringing MP3 files based on notifications from copyright holders.150
Identification of infringing providers is also a difficult issue, given that such
identifications may effectively be ex parte—in many cases, there’s no good
way for a content provider to learn of or respond to an accusation of
infringement. Also, the assessment of vicarious liability will involve some
close scrutiny into search business models; it appears that Google’s affiliate-
146. Some providers, both in the ISP and search engine contexts, have used such notices to
enforce non-copyright-related desires to suppress the material. See Online Policy Group v.
Diebold, Inc., 337 F. Supp. 2d 1195, 1198–1204 (N.D. Cal. 2004) (finding “material
misrepresentation” of infringement where posted “embarrassing” content was “not subject to
copyright protection”).
147. Joshua Urist, Who’s Feeling Lucky? Skewed Incentives, Lack of Transparency, and
Manipulation of Google Search Results Under the DMCA, 1 BROOK. J. CORP. FIN. &. COMM. L. 209,
219 (2006) (arguing that such Section 512(d) takedown requests should be publicly archived).
See Google, Digital Millennium Copyright Act, http://www.google.com/dmca.html (“Please
note that in addition to being forwarded to the person who provided the allegedly infringing
content, a copy of this legal notice may be sent to a third-party partner for publication and
annotation.”).
148. Perfect 10, Inc. v. CCBill LLC, 488 F.3d 1102, 1109–15, 1116–17 (9th Cir. 2007)
(discussing Section 512(d) and Section 512(i)).
149. The question of knowledge is relevant both in the standard for contributory
infringement and directly under Section 512; the ability to control infringers is relevant in
vicarious infringement and directly under Section 512.
150. See A&M Records v. Napster, Inc., 284 F.3d 1091, 1098–99 (9th Cir. 2002).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
36 93 IOWA LAW REVIEW [2007]
network advertising might be decisive in finding vicarious liability under the
right circumstances.151
Third-party trademark rights in content (as distinguished from
trademark rights in queries) are generally less dangerous for search engines.
Direct liability in trademark law for trademark use on providers’ pages seems
unlikely, given the trend in the keyword-advertising cases. Secondary liability
based on tests paralleling those in copyright might be argued by analogy to
the offline “swap meet” cases.152 eBay, the most like a swap meet of the major
search engines, has a rigorous trademark-protection policy and will take
down an auction based on a complaint from a trademark holder.153
2. Reputation
How different things are when the content flows are defamatory, rather
than infringing! Here, search engines are protected by Section 230 of the
Communications Decency Act, which gives any “provider . . . of an
interactive computer service” blanket immunity from being treated as the
“speaker of any information provided by another information content
provider or user.”154 Moreover, while search engines need not filter such
material, they are also immunized if they voluntarily, “in good faith,” remove
material that they believe to be “obscene, lewd, lascivious, filthy, excessively
violent, harassing, or otherwise objectionable.”155 Except in exceptional
circumstances, a party has no recourse against a search engine that facilitates
the distribution of personally harmful or defamatory content.156
151. See Perfect 10, Inc. v. Amazon.com, Inc., 487 F.3d 701, 711–12, 729–32 (9th Cir. 2007)
(noting that Google placed affiliate advertisements on sites allegedly hosting infringing
images).
152. See, e.g., Hard Rock Cafe Licensing Corp. v. Concession Serv., Inc., 955 F.2d 1143,
1148–50 (7th Cir. 1992) (discussing contributory and vicarious trademark liability for the
operator of a flea market where counterfeit goods are sold).
153. See eBay, How eBay Protects Intellectual Property (VeRO), http://pages.ebay.com/
help/tp/programs-vero-ov.html [hereinafter VeRO]. But see Complaint, Tiffany, Inc. v. eBay,
Inc., No. 04 CV 4607 (S.D.N.Y. June 18, 2004), available at http://cyberlaw.
stanford.edu/attachments/Tiffany%20eBay%20complaint.pdf (claiming that eBay is
secondarily liable for sales of trademarked goods). eBay is potentially more liable than most
stand-alone search engines under a vicarious-liability theory, since it has the ability to disable
any auction on its site. See VeRO, supra. Contributory trademark liability might be more broadly
generalizeable beyond eBay.
154. 47 U.S.C. § 230(c)(1)(2000).
155. Id. at § 230(c)(2)(A).
156. See Chicago Lawyers’ Comm. for Civil Rights Under the Law, Inc. v. Craigslist, Inc., 461
F. Supp. 2d 681, 698–99 (N.D. Ill. 2006) (applying Section 230 to preempt a claim under the
Fair Housing Act against an online classified-ads site and search engine for the discriminatory
housing ads posted by its users). For a remarkable example of the broadness of this immunity,
consider Sturm v. eBay, Inc., No. 1-06-CV-057926 (Cal. Super. Ct. July 27, 2006), in which eBay
was not required to remove defamatory feedback about a user, even when the user and the
defamer had entered into a settlement stipulating that the feedback was defamatory and both
had written eBay asking that it be removed. See Elise Ackerman, EBay Lawsuit Reveals Foibles of
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 37
Those exceptional circumstances might arise when the engine itself has
taken sufficient steps so that it could be identified as the provider of the
content. First, if it has encouraged the creation of the content and directed
its creation, it might be identified with the provider for liability purposes.157
This scenario is not a significant concern for a pure search engine, but the
growing integration of search engines with other applications raises
concerns, particularly for search engines associated with creative
communities.158 Second, to the extent that a search engine is viewed as a
speaker—something search engines are eager to encourage in the context
of defending themselves in suits over rankings159—its recommendations of
content potentially become endorsements of that content’s message.
Applicable precedents hold that services have substantial leeway to choose
which messages to pass along, but if the search engine itself adds some
content of its own to the recommendation (even, for example, a sentence
describing the linked-to page160), that additional content might fall outside
of Section 230’s protections.161
Site Feedback, SAN JOSE MERCURY NEWS, Aug. 9, 2006, at A1, available at
http://www.mercurynews.com/mld/mercurynews/news/local/15228670.htm.
Other countries have less categorical rules about intermediary immunity. See, e.g.,
Eiseres/Google, Rechtbank Amsterdam [Court Amsterdam], 30 April 2007, LJN BA3941
(Neth.), available at http://zoeken.rechtspraak.nl/resultpage.aspx?snelzoeken=true&
searchtype=ljn&ljn=BA3941&u_ljn=BA3941 (holding that Google is not obliged to prevent
unlawful information from entering its search results). Discussion of this case is available at Joris
van Hoboken, The Duty of Care of Search Engines (May 15, 2007), http://www.
jorisvanhoboken.nl/?p=32.
157. See Fair Housing Council of San Fernando Valley v. Roommate.com, LLC, 489 F.3d
921, 926 (9th Cir. 2007) (holding a roommate search service not immune under Section 230
for information supplied by users in drop-down menus on the service’s site). See generally Ken S.
Myers, Wikimmunity: Fitting the Communications Decency Act to Wikipedia, 20 HARV. J.L. & TECH.
163, 174–78 (2006) (discussing Section 230 cases and the extent to which a service can
encourage users to provide defamatory content without losing its immunity).
158. See, e.g., Chris Sherman, Puzzling Out Google’s Blogger Acquisition, SEARCH ENGINE
WATCH, Feb. 18, 2003, http://searchenginewatch.com/showPage.html?page=2161891
(discussing Google’s purchase of Pyra Labs, the creator of Blogger); Press Release, Google To
Acquire YouTube for $1.65 Billion in Stock, (Oct. 9, 2006), http://www.google.com/press/
pressrel/google_youtube.html (announcing that Google will acquire YouTube); Michael
Arrington, Windows Live Spaces Launches, Replaces MSN Spaces, TECHCRUNCH (Aug. 1, 2006),
http://www.techcrunch.com/2006/08/01/windows-live-spaces-launches-replaces-msn-spaces/.
159. See Search King, Inc. v. Google Tech., Inc., No. CIV-02-1457-M, 2003 U.S. Dist. LEXIS
27193, at *6 (W.D. Okla. May 27, 2003).
160. Yahoo’s directory describes each listed site in a sentence. Yahoo! Search Directory,
http://dir.yahoo.com/. The Open Directory Project does the same, with the taxonomy and
descriptions being supplied by volunteers. Open Directory Project, http://www.dmoz.org/.
161. Cf. Benjamin Cohen & Helen Nugent, Cole Tackles Google over Gay Link, TIMES
(LONDON), Mar. 7, 2006, at 31, available at http://www.timesonline.co.uk/tol/
news/tech_and_web/article738224.ece (explaining that Google search on “ashley cole”
returned “See results for: ashley cole gay” and describing Google spokesperson’s explanation
that alternative search suggestions are determined by computer algorithm).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
38 93 IOWA LAW REVIEW [2007]
The presence of search engines also reshapes the dynamics of the
struggle between defamer and defamee. Search engines can focus attention
on statements that previously would not have spread as far or as fast. Because
they selectively pair a query with results, they can firmly link a name to a
given piece of information. Indeed, precisely because people may wish to
search on others’ names, search engines regularly direct users to false
claims.162 Arguably, search engines also help diligent third parties discover
unflattering information about themselves before it has spread, allowing
them to move directly against the providers.163
In light of these dynamics, some scholars have argued that the law of
search should be modified (or interpreted) to provide at least some kind of
right of reply. By analogy to the Fair Credit Reporting Act, which allows
individuals to correct incorrect statements about their credit history (and
provides mechanisms to proceed both against original reporters and credit-
record search agencies), these scholars have suggested that search engines
should be required to respond to certain well-specified classes of
reputational harms. For example, the search engine could allow the subject
of an unflattering search result to annotate the result with an asterisked link
to her reply.164
The broad immunity search engines and other intermediaries currently
enjoy encourages reputational self-help.165 Scholars have argued that the
increasing democratization and interactivity of Internet communications
technologies mean that self-help should be the response of choice; one
confronted with an online falsehood should be encouraged first to
propagate the truth online.166 Given the enormous power that search
engines wield in shaping which messages are heard and which are not, it is
not obvious that the truth will necessarily be able to catch up with the
falsehood.167
162. For a domain-specific example, consider Don’t Date Him Girl, http://
dontdatehimgirl.com/, which allows users to search for personal reports posted by other users
about cheating men—by name. At least one alleged cheater has decided to take his chances
with Section 230 in a lawsuit against the site. See Carl Jones, Scorned Attorney Sues Kiss-and-Tell Web
Site, DAILY BUS. REV. (Florida), July 5, 2006, available at http://www.law.com/jsp/
article.jsp?id=1151658319991.
163. See Daniel Dasey, A Quick Self-Google Once a Day to Guard Your Reputation, SUN-HERALD
(Australia), May 23, 2004, available at http://www.smh.com.au/articles/2004/05/22/
1085176043551.html.
164. See Pasquale, supra note 3, at 135.
165. See generally R. Polk Wagner, On Software Regulation, 78 S. CAL. L. REV. 457 (2005)
(arguing that denying online actors legal recourse will encourage them to employ technical
self-help).
166. See, e.g., Edward A. Cavazos, Note, Computer Bulletin Board Systems and the Right of Reply:
Redefining Defamation Liability for a New Technology, 12 REV. LITIG. 231, 243–47 (1992).
167. See Cho & Roy, supra note 69 (arguing that web ranking mechanisms favor already
popular sites).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 39
There is another name for self-help directed at search engines: SEO.
Companies have been known to engage in substantial SEO to drive
unflattering messages about them from search engine prominence.168
Consider also “Googlebombing”: the process of creating hyperlinks using a
particular phrase and pointing to a particular page with the goal of
convincing search engines to return that page on a query for that phrase.
The most famous Googlebombs—George W. Bush as “miserable failure”;
Andy Pressman as “talentless hack”—have been widespread bottom-up
attempts to tarnish someone else’s reputation. (Google has since changed its
algorithms to make Googlebombs much harder to pull off.)169 Search
engines are just another media intermediary in age-old arms races to get
competing messages before the public.170
3. Privacy
Search engines give users remarkable ability to learn about others. They
can root out details that otherwise might have remained obscure and
correlate information from many different sources. They democratize
investigation, giving anyone the kind of power to develop a broad and deep
profile of their chosen target that previously would have required a
professional’s help.171
Search engines are uncomfortable with the privacy-invading power of
their own technologies. Google reacted with petulant anger to a CNet article
on the privacy concerns raised by search. The article included an extensive
profile of Google CEO Eric Schmidt, compiled by using Google, that
disclosed Schmidt’s political contributions, wife’s name, and hobbies.172
Google told CNet that it would refuse to talk to CNet reporters for a year.173
168. See Mark Glaser, Companies Subvert Search Results to Squelch Criticism, ONLINE
JOURNALISM REV., June 1, 2005, http://www.ojr.org/ojr/stories/050601glaser/.
169. See, e.g., Matt Cutts et al., A Quick Word About Googlebombs, OFFICIAL GOOGLE
WEBMASTER CENTRAL BLOG (Jan. 25, 2007), http://googlewebmastercentral.
blogspot.com/2007/01/quick-word-about-googlebombs.html (describing Google’s algorithmic
changes to end the effect of Googlebombs on its rankings).
170. For further discussion of Googlebombing, including these two examples, see infra Part
II.D.1.
171. See Randy Cohen, Is Googling O.K.?, N.Y. TIMES, Dec. 15, 2002, (Magazine), at 50
(discussing the ethics of Googling potential dates). Cf. Kevin Poulsen, MySpace Predator Caught by
Code, WIRED NEWS, Oct. 16, 2006, http://www.wired.com/science/discoveries/news/
2006/10/71948 (discussing the use of publicly available datasets to locate registered sex
offenders with Myspace profiles).
172. Elinor Mills, Google Balances Privacy, Reach, CNET NEWS, Aug. 3, 2005,
http://news.com.com/2102-1032_3-5787483.html.
173. Saul Hansell, Google’s Chief Is Googled, to the Company’s Displeasure, N.Y. TIMES, Aug. 8,
2005, at C4. The ban was dropped sub silentio within a few months. See Elinor Mills, Google to
Yahoo: Ours Is Bigger, CNET NEWS, Sept. 28, 2005, http://news.com.com/2102-1038_3-
5883345.html (quoting from a phone interview with Schmidt).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
40 93 IOWA LAW REVIEW [2007]
The legal baseline when it comes to search subjects’ privacy, as with
defamation, is that search engines cannot be held liable for the information
they pass along. But on a policy level, third-party-privacy problems are even
thornier than user-privacy problems.174 Users and search engines have a
direct relationship that allows privacy-sensitive users to negotiate for better
protections (at least in theory). Third parties stand in no such relationship
to search engines.175 The paradigmatic nightmare case for third-party
privacy—the homicidal stalker—is a lot worse than the corresponding
nightmare cases for search users—embarrassment for highly sensitive
searches.
One interesting idea of the relationship between user and third-party
privacy is that there should be some rough symmetry. That is, informational
tools should disclose to search subjects that others are searching for
information about them, possibly even including the names of the searchers.
At least within closed online communities, social norms often favor some
rough balance of privacy reciprocity between searchers and searchees.176
Third-party privacy in the age of search raises some important structural
problems that set this issue apart from the intellectual-property and
defamation problems discussed above. The chaotic state of privacy law
means that primary liability for providers who release privacy-damaging
information is sometimes available and sometimes not. Trade secrets or
medical information? Well-defined causes of action will lie.177 Your address
and phone number? Probably not. One reason for this patchwork approach
to primary liability is that privacy harms tend to stem from the aggregation
of information, rather than from any single piece. Search engines and other
information-location tools turn what would formerly have been
inconsequential disclosures into the component parts of a genuine privacy
violation.
174. To be analytically complete, we should consider provider privacy. In practice, however,
provider privacy is not much of a problem. Providers who do not wish to be found by searching
usually opt off the public Internet entirely. Accidental leaks of information found by search
engines raise issues similar to those raised by third-party privacy concerns. That said, a provider
that accidentally releases information harmful to itself is almost certainly the least-cost avoider
as compared with a search engine.
175. Self-help is not entirely out of the question. Douglas Coupland’s novel JPod includes a
character who creates a juicy but false web site about herself (stating that she lost hundreds of
pounds on the Subway diet but has cheated by sneaking junk food snacks), partly for
amusement, but partly to throw searchers off the track of true information about her. DOUGLAS
COUPLAND, JPOD 118–21, 166 (2006).
176. See Lior Strahilevitz, Friendster and Symmetrical Privacy, UNIV. OF CHI. L. SCH. FACULTY
BLOG (Oct. 6, 2005), http://uchicagolaw.typepad.com/faculty/2005/10/friendster_and_.html.
177. See, e.g., UNIFORM TRADE SECRETS ACT § 1(b)(2)(b)(ii) (defining “misappropriation”
to include disclosure of information by one who acquired it “under circumstances giving rise to
a duty to maintain its secrecy or limit its use”); McCormick v. England, 494 S.E.2d 431, 437 (S.C.
Ct. App. 1997) (holding that “an actionable tort lies for a physician’s breach of the duty to
maintain the confidences of his or her patient”).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 41
It’s also worth discussing the relative fault of the user and the third
party. In the classic third-party intellectual-property case, the provider and
user are joint tortfeasors, collaborating to rook the innocent intellectual-
property holder. In the classic defamation case, the user is a deluded
innocent, the third party is a besmirched innocent, and the provider is a
dirty liar.178 In a real case of privacy violation, though, the user may
sometimes be the most blameworthy party: she’s the one who set out to
build a dossier on her subject. Interestingly, the subject may have more self-
help power to prevent a privacy spill than a drive-by defamation. Anyone can
make up lies about me pretty much no matter what I do, but by staying off
the grid and keeping others from learning much about me, I can
(imperfectly) mitigate the risks of my private information becoming widely
known. That said, the choice to remain offline is less and less available, and
one may have no ability to opt out of many data flows.179
4. User Virtue
Search censorship is on the rise around the world as more and more
governments require search engines to filter results or block certain
keywords from being searched.180 The common thread uniting these various
governmental demands is concern for what can best be described as their
citizens’ “virtue”: governments fear that some content will corrupt any users
who see it. That corruption can involve personal morality, as with
pornography and graphic violence; it can also involve political morality, as
with hate speech and dissent. Debates over whether these information flows
should be suppressed are inherently political and deeply controversial.
178. Some users seek out negative information on others, it’s true, but it’s still hard to see
those users as guilty of anything more than having an uncharitable attitude toward their fellow
man. If the information is true, nothing wrong has taken place. If the information is false and
the users know it’s false, it may no longer be genuinely defamatory; all that remains is a kind of
residual tarnishment. Additionally, if the information is false and the users think it’s true, the
deceit is the provider’s fault. Only when a user republishes the libel—and thus becomes a
provider herself—does her culpability ripen into something worthy of the name.
179. See, e.g., Ryan Singel, Sex Lube Maker’s 250K Customer List Slides onto Net, THREAT LEVEL
(Apr. 24, 2007), http://blog.wired.com/27bstroke6/2007/04/sex_lube_makers.html
(describing how the combination of poor security at Astroglide’s web site and searchability
means that Google searches on names of customers now return the name of the product they
ordered).
180. Only rarely do nongovernmental actors even try to force search censorship, since they
rarely have legal standing to do so. The most notable American exception is Complaint, Toback
v. Google, Inc., No. 06-007246 (N.Y. Sup. Ct. May 4, 2006), available at http://
casedocs.justia.com/new-york/nyedce/2:2006cv02692/257132/1/1.pdf, a private suit in
nuisance and intentional infliction of emotional distress that accused Google of allowing traffic
in child pornography. The causes of action were obviously preempted by Section 230, and the
plaintiff, a New York state legislator, dropped it within two months, after Google “offered to sit
down and discuss the issues.”
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
42 93 IOWA LAW REVIEW [2007]
The application of different national legal restrictions to search181
mirrors one of the oldest and most contentious debates in Internet law:
whether ISPs must, may, or must not filter out dangerous traffic flowing
through their networks. Different national cultures and laws have treated
different forms of Internet speech as dangerous. Child pornography touches
a nerve in the United States; in France, it is hate speech; in China, it is
criticism of the government. Government-mandated filtering and blocking
at the network layer is not new. In the United States, a Pennsylvania law that
would have required ISPs to filter content was declared unconstitutional;182
the Chinese government engages in massive technological censorship
through control over the routers that connect Chinese portions of the
Internet to the rest of the world.183
The extension of such requirements to search engines is not exactly
new, either. In 2000, French groups discovered that Yahoo!’s chat rooms
and auctions included neo-Nazi material and ultimately won a court order
requiring Yahoo! to block French access to such auctions, as well as “to any
other site or service that may be construed as constituting an apology for
Nazism or a contesting of Nazi crimes.”184 Yahoo!’s attempt to block United
States enforcement of the order was ultimately inconclusive. Long before
the Ninth Circuit dismissed Yahoo!’s declaratory judgment action, Yahoo!
had decided that it preferred to do business in France on French terms and
came into compliance with the order.185 This has been the pattern for many
United States technologies companies doing business abroad ever since—
initial protestations, followed by compliance with local law.186
Today, China is the most aggressive search-blocking regime.187 Search
engines—like most other Internet intermediaries—are expected to block a
181. See CenSEARCHip, About CenSEARCHip, http://homer.informatics.indiana.edu/
censearchip/about.html (describing a service that compares search results in China, France,
Germany, and the United States); see also JACK GOLDSMITH & TIM WU, WHO CONTROLS THE
INTERNET? 149 (2006) (“[W]hat we once called a global network is becoming a collection of
nation-state networks . . . .”).
182. See Cen. for Democracy and Tech. v. Pappert, 337 F. Supp. 2d 606, 663 (E.D. Pa. 2004)
(holding unconstitutional a Pennsylvania anti-pornography ISP censorship law).
183. See Richard Clayton et al., Ignoring the Great Firewall of China 1 (2006) (unpublished
manuscript), available at http://www.cl.cam.ac.uk/~rnc1/ignoring.pdf (describing technologies
used to prevent Chinese access to disapproved Internet sites and techniques for evading this
censorship).
184. See Yahoo! Inc. v. La Ligue Contre Le Racisme et L’Antisemitisme, 433 F.3d 1199, 1202
(9th Cir. 2006) (providing an English translation of the French order).
185. Id. at 1215–18. The en banc court split: some judges would have allowed the suit to go
forward, some would have dismissed for lack of personal jurisdiction, and some would have
dismissed for lack of ripeness. A majority of the en banc panel voted for dismissal, but neither
reason for dismissal commanded a majority on its own.
186. See GOLDSMITH & WU, supra note 181, at 159–60.
187. See Thompson, supra note 63 (describing the Chinese government’s Internet
censorship laws and Google’s operations in the country).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 43
slightly nebulous set of forbidden content (the vagueness of the standards
leads to inconsistent and unpredictable overblocking).188 Falun Gong,
Tibetan culture, and political dissidence are the principal, but by no means
the only, targets.189 These policies—particularly when adopted by “do no
evil” Google—have led to a domestic political backlash, including the
introduction in Congress of the Global Online Freedom Act, which would
require search engines to refuse “Internet-restricting countr[ies’]”
governments’ requests to alter results and would require search engines to
notify the United States government of their blocklists.190 The complexity of
varying international standards and the concern that compliance with local
law may lead to human-rights violations have led some companies to ask the
United States government for help.191
Governments have also learned that they can ask or force
intermediaries to identify Internet users who exchange forbidden content.
For example, Chinese dissidents convicted and imprisoned with information
that Yahoo! supplied have sued Yahoo! in the United States under the Alien
Tort Claims Act.192 Search companies have been inconsistent in their
willingness to contest demands for private information from local
authorities. Google strongly resisted a Department of Justice subpoena for
query data in Gonzales v. Google,193 but it has reached arrangements with
authorities in Brazil to identify users of its Orkut social-networking service
who are accused of spreading child pornography or engaging in hate
speech.194
Government interests in censoring or finding the authors of unwanted
speech are in direct tension with many user and provider interests in
conducting searches. Blocking is a deliberate and heavy-handed form of
bias. Government recruitment of search also directly threatens both user
and provider privacy. On the user side, search queries—whether obtained
from the search engine or from other sources—can be highly incriminating
188. This self-censorship system is not unique to China; search providers in Germany use a
similar system to suppress hate speech. See Subcode of Conduct for Search Engine Providers of
the Association of Voluntary Self-Regulating Multimedia Service Providers, http://
www.fsm.de/en/SubCoC_Search_Engines.
189. See, e.g., Oxblood Ruffin, Google, China, and Genocide, CULT OF THE DEAD COW
COMMUNICATIONS, Apr. 22, 2007, http://www.cultdeadcow.com/cDc_files/cDc-0409.php
(describing Google’s Chinese-search censorship as participation in “cultural genocide”).
190. H.R. 4780, 109th Cong. §§ 202(1), 203 (2d Sess. 2006).
191. See Anne Broache, Web Giants Ask for Feds’ Help on Censorship, CNET NEWS, Jan. 31, 2007,
http://news.com.com/2102-1028_3-6154930.html.
192. See Complaint of Tort Damage, Wang Xiaoning v. Yahoo! Inc., No. 07-2151 (N.D. Cal.
Apr. 19, 2007).
193. Gonzales v. Google, Inc., 234 F.R.D. 674, 680 (N.D. Cal. 2006) (“Google primarily
argues that the information sought by subpoena is not reasonably calculated to lead to evidence
admissible in the underlying litigation and that the production of information is unduly
burdensome.”).
194. Broache, supra note 48.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
44 93 IOWA LAW REVIEW [2007]
evidence.195 And on the provider side, search provides another tool in the
government’s surveillance toolkit. Unlike with most private surveillance, the
government may be interested in searching entire populations for activity it
considers suspicious. Think of the National Security Agency (“NSA”)
surveillance program, which amounts to a gargantuan search project across
the huge daily volume of phone calls, apparently seeking patterns of calls
fitting specified profiles.196
D. SEARCH ENGINES’ INTERESTS
We finish this survey with a discussion of search engines’ interests. The
basic search business model may be summed up quite simply: Provide high-
quality results to attract users, sell ads, and rake in the bucks. This Section
discusses three significant threats to this business model: SEO, click fraud,
and competition. SEO manipulators and click fraudsters pose an operational
threat, sucking the value out of results. Contract and commercial-fraud law
provide protections from these sorts of dishonesty. Other search engines
pose a competitive threat. Intellectual-property law and antitrust law both
provide protection for the needs of competing search engines.
1. Preventing Search Engine Optimization
SEO can be described as a deliberate attempt by a provider to introduce
results bias against its competitors. As such, it harms users, whose searches
become less useful, as well as legitimate providers, whose content becomes
less visible. Strong market incentives compel search engines to combat SEO
because by doing so, engines can give users better results. The consequence
is a technical arms race between engines and manipulators. Search engines
jealously guard their ranking algorithms as a way of maintaining an edge in
this race.197 Because search results generally are public, however,
manipulators can engage in extensive reverse engineering. Even if they do
not know precisely how the engine computes results, they can create
approximate models of the sorts of content it favors and disfavors.198
195. See United States v. Schuster, 467 F.3d 614, 617 (7th Cir. 2006) (upholding factual
findings based in part on Google searches by the defendant); K.C. Jones, Murder Suspect’s Google
Searches Spotlighted in Trial, CHANNEL WEB NETWORK, Nov. 11, 2005, http://www.crn.com/it-
channel/173602157 (discussing a prosecutor’s claim that a murder defendant used Google to
learn the “depth and topography of a lake where the body of his wife . . . was found”).
196. Press Release, President George W. Bush, President Bush Discusses NSA Surveillance
Program (May 11, 2006), available at http://www.whitehouse.gov/news/releases/
2006/05/20060511-1.html.
197. See, e.g., Matt Cutts, Notifying Webmasters of Penalties, GADGETS, GOOGLE, AND SEO (Apr.
26, 2006), http://www.mattcutts.com/blog/notifying-webmasters-of-penalties/ (discussing the
tradeoffs involved in notifying providers that their sites are being penalized).
198. See, e.g., Dave Tiberio, Reverse-Engineering Search Engine Ranking Algorithms,
WEBPRONEWS.COM, http://www.webpronews.com/node/23170/print (explaining the basics of
reverse engineering search engine algorithms).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 45
Pure SEO has not generated much in the way of litigation. Because
being caught engaging in SEO is the most common way to be stricken from
an engine’s index, the question of whether a provider is engaging in SEO
has sometimes been raised as a collateral issue in placement suits.199 Search
engines use their operational need to prevent SEO as an argument against
judicial oversight of their ranking decisions. Although it’s not inconceivable
that providers and users upset at fraud-heavy results might sue search
engines, these parties are in the same boat as engines. Engines almost never
make specific promises to stamp out search fraud, and there is little more
that most engines could do.
It is quite possible that the next few years will see some lawsuits against
providers that allege the use of SEO tactics. The trademark theories
mentioned above were originally developed in direct suits by trademark
holders (i.e., other content providers) against alleged manipulators.200
Courts have recognized that some techniques of content design are
deceptively manipulative and cause harm to legitimate providers, and it is
possible that innovative pleading could properly state other business torts
against manipulators. Similarly, luring users to one’s content through SEO
raises significant false-advertising concerns. In these cases, competitors,
users, and consumer-protection agencies might all be proper plaintiffs.
One limit to these theories is that they see the harm as transactions
facilitated or foiled by the misdirection. SEO techniques, however, are
neither necessarily commercial nor necessarily centralized. Consider again
Googlebombing.201 The most famous Googlebomb of all time may be the
linking of the phrase “miserable failure” to the White House biography of
President George W. Bush.202 Similarly, companies have learned how to
engage in “sock puppetry,” creating fake content and personae to express
their point of view and suppress negative opinions in search results.203 The
misrepresentations, if any, involved in these techniques are not easily
199. See generally, e.g., Search King, Inc. v. Google Tech., Inc., No. CIV-02-1457-M, 2003 U.S.
Dist. LEXIS 27193 (W.D. Okla. May 27, 2003).
200. See generally Brookfield Commc’ns, Inc., v. W. Coast Entm’t Corp., 174 F.3d 1036 (9th
Cir. 1999).
201. See Adam Mathes, Filler Friday: Google Bombing, ÜBER.NU, Apr. 6, 2001, http://uber.
nu/2001/04/06/.
202. See “Miserable Failure” Links to Bush, BBC NEWS, Dec. 7, 2003, http://
news.bbc.co.uk/2/hi/americas/3298443.stm (describing the “Miserable Failure”
Googlebomb). Whether the critics of George W. Bush who created this Googlebomb will feel as
proud of it when the White House has its next occupant is another matter entirely.
203. See, e.g., Brad Stone, The Hand That Controls the Sock Puppet Could Get Slapped, N.Y.
TIMES, July 16, 2007, at C1, available at http://www.nytimes.com/2007/07/16/technology/
16blog.html; see also Glaser, supra note 168; Wikipedia, Sock Puppetry, http://en.wikipedia.org/
wiki/Wikipedia:Sock_puppet (“Use of sock puppets is discouraged in most cases.”).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
46 93 IOWA LAW REVIEW [2007]
characterized. Opinions differ as to whether search engines should try to
suppress Googlebombs and sock puppets.204
2. Preventing Click Fraud
Click fraud has led to more litigation,205 probably because it involves
both money taken directly from advertisers’ pockets and the contractual
relationship between advertisers and engines. The Coase theorem suggests
that systematic click fraud will simply be reflected in lower prices for
advertising, since click fraud increases the number of clicks charged per sale
generated.206 Competitors who target particular advertisers for fraudulent
purposes present a more difficult problem. One might expect that
advertisers would demand clauses relieving them of the responsibility of
paying for fraudulent clicks if they are targeted.207 Enforcing such clauses,
however, requires that advertisers be able to monitor the clicks for which
they are charged.208 Search engines have been reluctant to share the
204. See generally Clifford Tatum, Deconstructing Google Bombs: A Breach of Symbolic Power or Just
a Goofy Prank?, FIRST MONDAY, Oct. 2005, http://www.firstmonday.org/issues/
issue10_10/tatum/index.html (suggesting that attempts at suppression will have little impact).
Compare Saul Hansell, Foes of Bush Enlist Google to Make Point, N.Y. TIMES, Dec. 8, 2003, at C8
(“We just reflect the opinion on the Web . . . .”), with Danny Sullivan, Googlebombing Now a
“Prank” and Not Web’s Opinion, Says Google, SEARCH ENGINE WATCH (Sept. 19, 2005), http://blog.
searchenginewatch.com/blog/050919-095321 (arguing that Google has changed its views about
the nature of Googlebombing).
205. See Class Action Complaint and Jury Trial Demand, Crafts by Veronica v. Yahoo!, Inc.,
No. 2:06-cv-01985 (D.N.J. May 1, 2006); Class Action Complaint and Jury Trial Demand,
Draucker Dev. v. Yahoo! Inc., No. CV06-2737 (C.D. Cal. May 4, 2006); Final Order and
Judgment Approving Settlement, Lane’s Gifts and Collectibles, Inc. v. Yahoo! Inc., No. CV-2005-
52-1 (Ark. Cir. Ct. July 26, 2006), available at http://blog.ericgoldman.org/
archives/lanegiftsac.pdf; Findings and Preliminary Order Approving Settlement of Class Action
and Directing the Issuance of the Notice to the Class, Checkmate Strategic Group, Inc. v.
Yahoo! Inc., No. 2:05-CV-04588-CAS-FMO (C.D. Cal. June 28, 2006); Advanced Internet Tech.,
Inc. v. Google, Inc., 2006 WL 889477 (N.D. Cal. stay entered Apr. 5, 2006); Complaint, CLRB
Hanson Industries LLC v. Google, Inc., No. 1-05-CV-046409 (Cal. Sup. Ct. Aug. 3, 2005); see also
David A. Vise, Clicking to Steal, WASH. POST, Apr. 17, 2005, at F01 (describing the lawsuit by
Google against Auctions Expert alleging click fraud).
206. Some have argued that search engines have no incentive to police click fraud because
they can charge for illegitimate clicks. See, e.g., Brian Grow & Ben Elgin, Click Fraud: The Dark
Side of Online Advertising, BUSINESSWEEK, Oct. 2, 2006, at 46, 51 (quoting Martin Fleischmann,
an Internet entrepreneur and victim of click fraud, “‘I told Yahoo years ago,’ he says, ‘“If this
[fraudulent clicks] was costing you money instead of making you money, you would have
stopped this.”’”). Not so. The Coasean exchange returns to search engines the necessary
incentive, as they will be able to increase their cost-per-click if they reduce fraud rates.
207. But see, e.g., Yahoo! Search Marketing, Click Fraud FAQ, http://searchmarketing.
yahoo.com/legal/clickfraud.php (“[E]ven though we are not obligated to, we voluntarily designed
the Click Protection System” to identify click fraud (emphasis added)).
208. See generally GOOGLE, INC. CLICK QUALITY TEAM, HOW FICTITIOUS CLICKS OCCUR IN
THIRD-PARTY FRAUD AUDIT REPORTS 4 (2006), http://www.google.com/adwords/
ReportonThird-Party
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 47
necessary data with advertisers, fearing that it would permit reverse
engineering of the search engine’s fraud-detection algorithms.209
Conversely, overly zealous enforcement of anti-fraud policies upsets
advertisers, as the penalties unilaterally imposed by search engines against
sources of click fraud include banishment from search advertising.210
Independent auditing of click counts and of anti-fraud programs may be the
wave of the future.211
Click fraud, because of its links to other ugly online practices,212 has also
embroiled some search engines in litigation involving those links.213
Consumers and state attorneys general have begun to sue spyware makers
and the advertising networks linked to them.214 Following the chain of
business relationships leads inevitably back to the search engines whose
affiliate networks provide the ad delivery and billing at the heart of many of
these schemes.215 Those who distribute spyware also do so through
advertisements placed on search engine sites (and on sites highly ranked
due to SEO), suggesting that search engines (which often forbid these
tactics in their advertiser guidelines) may bear some moral and legal
responsibility for the problems created by spyware.216 Their central
ClickFraudAuditing.pdf (arguing that third-party click-fraud-auditing firms overestimate click-
fraud rates).
209. See ALEXANDER TUZHILIN, THE LANE’S GIFTS V. GOOGLE REPORT, http://
googleblog.blogspot.com/pdf/Tuzhilin_Report.pdf (describing Google’s reasons for
withholding details of click fraud from advertisers). This monitoring problem creates a close
connection between search engines’ two biggest operational problems: SEO and click fraud.
210. See generally BATTELLE, supra note 1.
211. See Press Release, Interactive Adver. Bureau, The Interactive Industry Commits to the
Development of Click Measurement Guidelines (Aug. 2, 2006), available at
http://www.iab.net/news/pr_2006_08_02.asp (announcing an industry-wide group to develop
a set of Click Measurement Guidelines that would standardize how invalid and fraudulent clicks
are defined and counted).
212. See Damien Cave, The Parasite Economy, SALON, Aug. 2, 2001,
http://archive.salon.com/tech/feature/2001/08/02/parasite_capital/index.html (coining the
term); Stefan Gorling, An Introduction to the Parasite Economy, in EICAR 2004 CONFERENCE
PROCEEDINGS (U.E. Gattiker ed., 2004), available at http://papers.ssrn.com/sol3/
papers.cfm?abstract_id=683081 (generalizing it).
213. See, e.g., Crafts by Veronica v. Yahoo!, Inc., No. 2:06-cv-01985 (D.N.J. May 1, 2006).
214. See, e.g., Complaint for Injunctive and Additional Relief, Washington v. Secure
Computer, LLC, No. C-06-0126 (W.D. Wash. filed Jan. 24, 2006); Notice of Verified Petition,
New York v. DirectRevenue, LLC (N.Y. Sup. Ct. filed Apr. 4, 2006); Settlement Agreement and
Limited Release, Sotelo v. DirectRevenue, LLC, No. 05-C-2562 (N.D. Cal. Mar. 10, 2006).
215. See Ben Edelman, The Spyware-Click-Fraud Connection—and Yahoo’s Role Revisited (Apr. 4,
2006), http://www.benedelman.org/news/040406-1.html (describing the role of Yahoo!’s
affiliate network in placing advertisements in spyware). Edelman is co-counsel for plaintiffs in
Crafts by Veronica v. Yahoo! Inc., No. 2:06-cv-01985 (D.N.J. May 1, 2006).
216. See Ben Edelman, Pushing Spyware Through Search (Jan. 26, 2006),
http://www.benedelman.org/news/012606-1.html (analyzing the prevalence of spyware-laden
sites in Google results and advertisements for some common searches).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
48 93 IOWA LAW REVIEW [2007]
placement athwart many information flows may make search engines the
least-cost avoiders for some of these ecological problems.217
3. Innovation
Search engines compete with other search engines for users.218 They
compete by offering more complete indices of the Internet, by providing
more responsive results, and by integrating their searches with other
features valued by users. (Their actual revenue source, advertising, is
dependent on their ability to attract users, so market share by searches is a
good indicator of competitive success.) All three of these techniques involve
a sometimes-frenetic pace of innovation. Search engines zealously wield
intellectual-property rights to protect their innovations from being
appropriated.219
Search engines primarily use trade-secret techniques. They closely
guard their ranking and indexing algorithms and routinely invoke the need
to protect this secrecy in litigation that might expose operational details.220
In practice, this secrecy is incomplete because the public disclosure of
results permits limited reverse engineering. Search engines have also
discovered the value of allowing nearly unlimited usage of their search
facilities, including through automated application programming interfaces
(“APIs”);221 they therefore are reluctant to take obfuscatory technical steps
against reverse engineering. Further, the pressures of public relations
encourage search engines to trumpet various advances and tweaks to their
algorithms, if only in general terms.222
217. On search engines’ centrality and connections to web spam, see YI-MIN WANG ET AL.,
SPAM DOUBLE-FUNNEL: CONNECTING WEB SPAMMERS WITH ADVERTISERS (2007), available at
http://www.cs.ucdavis.edu/~hchen/paper/www07.pdf.
218. Currently, according to comScore and Nielsen NetRatings, Google is the market
leader with a roughly fifty-percent share; Yahoo! has about twenty-five percent, and MSN
another ten percent. AOL and Ask also have significant shares of perhaps five percent each,
and the remainder is split among a great many minor search engines. See Press Release,
comScore, comScore Releases June U.S. Search Engine Rankings (July 16, 2007), available at
http://ir.comscore.com/releasedetail.cfm?ReleaseID=254477; Press Release,
Nielsen//NetRatings, Nielsen//NetRatings Announces May U.S. Search Share Rankings (June
20, 2007), http://www.nielsen-netratings.com/pr/pr_070620.pdf.
219. Jonathan Thaw & Susan Decker, Google to Subpoena Yahoo, Microsoft on Book Scanning,
BLOOMBERG MARKETS, Oct. 5, 2006, http://www.bloomberg.com/apps/news?pid=20601103&
sid=amfuMLMq__H8.
220. See Gonzales v. Google, Inc., 234 F.R.D. 674, 684 (N.D. Cal. 2006) (considering
Google’s allegation that revealing random URLs and user queries would expose trade secrets).
221. See Robert D. Hof, Mix, Match, and Mutate, BUSINESSWEEK, July 25, 2005, at 72, available
at http://www.businessweek.com/magazine/content/05_30/b3944108_mz063.htm (describing
“mash-up” applications made by combining multiple programmatic APIs). Such mash-ups
would be nearly impossible if search engines and other web companies treated their results as
resources to be closely guarded.
222. See, e.g., Cutts et al., supra note 169.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 49
Search engines also take appropriate measures to guard the protected
status of their trade secrets. They routinely require employees to sign
nondisclosure and noncompetition agreements to prevent the departure of
search secrets to rivals.223 They typically require visitors to sign
comprehensive nondisclosure agreements, as well.224 Secrecy at Google, in
particular, is almost a way of life.225
Google’s patent on PageRank is best-known,226 but all of the major
search engines have patent portfolios. They aren’t afraid to use them, either.
In the run-up to its IPO, Google settled, for stock worth approximately $300
million, an outstanding patent-infringement lawsuit brought by Yahoo!. The
suit involved a patent on the technique of displaying keyword-triggered
search ads based on the bids submitted by advertisers—the technique at the
heart of most search advertising today.227 A steady drumbeat of search
engine patent suits continues.228
Search engines also use copyright and trademark to protect their
business models, although these matters are less litigated. Search engines
possess valid copyrights in their software and interfaces; they have valuable
223. See, e.g., Google, Inc. v. Microsoft Corp., 415 F. Supp. 2d 1018, 1020 (N.D. Cal. 2005)
(discussing Microsoft’s attempt to prevent Kai-Fu Lee, former vice president at Microsoft, from
becoming vice president at Google).
224. See Nick Denton, Google: This NDA Never Existed, VALLEYWAG (Jan. 22, 2007),
http://valleywag.com/tech/google/this-nda-never-existed-230407.php (reprinting Google’s
standard nondisclosure agreement).
225. See Robin Sidel et al., At Google, Mum’s the Word About Almost Everything, WALL ST. J., Apr.
27, 2007, at B1 (detailing Google’s secrecy with both investors and the outside world).
226. U.S. Patent No. 7,058,628. The patent is “owned by Stanford University, but licensed
exclusively to Google until 2011.” BATTELLE, supra note 1, at 130.
227. See George Mannes, Yahoo! Gets Bigger Stake in Google, THESTREET.COM, Aug. 9, 2004,
http://www.thestreet.com/pf/tech/georgemannes/10177217.html.
228. See generally Hyperphrase Techs., LLC v. Google Inc., No. 06-C-199-5, 2006 U.S. Dist.
LEXIS 64918 (W.D. Wis. Sept. 7, 2006); Skyline Software Sys. v. Keyhole, Inc., 421 F. Supp. 2d
371 (D. Mass. 2006), summary judgment granted to defendant Google, Inc., No. 06-10980-DPW, 2007
U.S. Dist. LEXIS 16053 (D. Mass. March 7, 2007); Netjumper Software, L.L.C. v. Google, Inc.,
No. 04-80366-CV, 2005 U.S. Dist. LEXIS 27813 (S.D.N.Y. Nov. 10, 2005); FindWhat.com v.
Overture Servs., Inc., No. 02 Civ. 447 (MBM), 2003 U.S. Dist. LEXIS 2450 (S.D.N.Y. Feb. 13,
2003).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
50 93 IOWA LAW REVIEW [2007]
trademark and trade-dress rights in their brands.229 Whether search results
are copyrightable as such is a debatable proposition.230
4. Competition
Any individual search engine would love to dominate the market for
search. But they share a collective interest that the market remain
competitive. The law of unfair competition—unconnected with
monopolization as such—provides a baseline of legitimate and illegitimate
business practices in the search market.231 Antitrust law hasn’t had much to
say about search, but that may change. Currently, Google’s fifty-percent
market share makes it the big fish among major search engines, and the
obvious first target for antitrust concern. The overall web search market is
dominated by the top few providers.
Are these conditions alarming? Arguments can be made both ways. On
one hand, users can easily switch search engines if they are unhappy with a
particular engine’s practices. Search engines are trying to change that with
increased personalization, but so far, little prevents users from switching.
The greater competition concern may come from the rising costs of entry;
the Internet continues to grow rapidly, and the SEO arms races have meant
that sophisticated and computationally expensive algorithms appear to be
part of the price of offering useful search. The real question may be whether
one considers near technological neighbors to be good substitutes for
centralized search. Reclassifying various technologies—e.g., del.icio.us’s
social bookmarks or eBay’s product search—as “search” would greatly
increase the denominator and reduce relative market shares. These near
neighbors may also have lower barriers to entry than server-farm-heavy,
centralized search.
229. See Penelope Patsuris, The Making of a $2 Billion Brand, FORBES.COM, Feb. 21, 2003,
http://www.forbes.com/2003/02/21/cx_pp_0221google.html (describing the ubiquity of the
word “google”); see also Frank Ahrens, So Google Is No Brand X, but What Is ‘Genericide’?, WASH.
POST, Aug. 5, 2006, at D01, available at http://www.washingtonpost.com/wp-
dyn/content/article/2006/08/04/AR2006080401536.html (describing a baseless attempt by
Google trademark attorneys to object to a newspaper’s statement that “Google” was in common
use as a generic verb); Beth Lipton Kriegel, Yahoo Not Amused by Pot Parody Site, CNET NEWS, Jan.
11, 1999, http://news.com.com/2100-1023-219986.html (describing Yahoo!’s action against
yahooka.com).
230. See generally Dan Burk, Method and Madness in Copyright Law, 2007 UTAH L. REV.
(forthcoming 2007), available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=999433
(discussing the idea-expression dichotomy and the copyrightability of the results of automated
processes).
231. Cf. Doug Young, Yahoo and Former China Head in Brewing Legal Battle, REUTERS, Aug. 17,
2006, available at http://services.inquirer.net/print/print.php?article_id=15801 (noting that
Qihoo sued Yahoo for defamation), with Sumner Lemon, Yahoo China Sues Rival Portal, IDG
NEWS SERV., Sept. 29, 2006, available at http://www.infoworld.com/article/06/09/
29HNyahoochinasues_1.html?PORTALS (noting that Yahoo sued Qihoo for unfair
competition).
GRIMMELMANN_FINAL.DOC 11/21/2007 12:05 PM
THE STRUCTURE OF SEARCH ENGINE LAW 51
Some search-ranking lawsuits filed by disappointed providers have
raised antitrust theories, although they have failed to explain why the poor
ranking of a provider constitutes an antitrust injury.232 Google’s proposed
purchase of web-advertising powerhouse DoubleClick has raised both
eyebrows and concerns.233 Where market power in a related market—e.g.,
desktop operating systems or provision of broadband telecommunications—
is more clearly established, search engines have appealed to the law and
policy of antitrust to prevent that power from being leveraged into the
search market. Thus, competing search engines have objected to the
integration of a search engine into other applications—e.g., through a
search box in a browser—but have not convinced antitrust regulators that
the practice is worthy of concern.234 Similarly, search engines have appealed
to telecommunications regulators, asking that network operators be given
little market power to discriminate among different traffic flows across their
networks.235 Search engines fear that network operators would use this
power to extract rents from them.236
III. INTERCONNECTIONS IN SEARCH ENGINE LAW
Already, some of the connections and conflicts among these problems
in search engine law should be evident. This Part examines five in more
depth. In each case, a broad view of search—one that considers the
interrelationships among many different doctrines—provides a clearer and
more helpful analysis of what is really at stake in a given legal dispute.
First, many forms of relief against search engines are functional
substitutes for one another. Some of the hardest-fought issues in search
policy are all but moot in light of doctrines from other areas. In general,
such doctrinal distinctions are unstable; the broad view of search forces us to
recognize that the technical centrality of search engines puts strains on
many different areas of law.
Second, search engines raise problems of unaccountable discretion.
Two natural tools to investigate and remedy abuses of discretion are greater
disclosure of the basis for ranking decisions and mandated corrections to fix
232. See Person v. Google, Inc., 456 F. Supp. 2d 488, 491–92 (S.D.N.Y. 2006) (alleging that
Google does not allow smaller customers to purchase certain AdWords but dismissing on other
grounds); Tushnet, supra note 75 (characterizing KinderStart’s allegations of Google antitrust
violations as “incomprehensible”).
233. Steve Lohr, Google Deal Said to Bring U.S. Scrutiny, N.Y. TIMES, May 29, 2007, at C1,
available at http://www.nytimes.com/2007/05/29/technology/29antitrust.html.
234. Ina Fried, Vista Search Seems Fair, Regulators Say, CNET NEWS, May 12, 2006,
http://news.com.com/2102-1014_3-6071851.html.
235. See Eric Schmidt, A Note to Google Users on Net Neutrality,
http://www.google.com/help/netneutrality_letter.html (asking concerned citizens to voice
their support for net neutrality).
236. Arshad Mohammed, Verizon Executive Calls for End to Google’s “Free Lunch”, WASH. POST,
Feb. 7, 2006, at D1.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
52 93 IOWA LAW REVIEW [2007]
misleading rankings. But these tools run squarely up against search engines’
operational interests in fighting search fraud, against their competitive
interest in trade secrecy, and perhaps even against user interests in privacy.
Taking a broad view of search forces us to weigh the costs and benefits of
these remedies carefully.
Third, user privacy concerns must be understood in the context of
search engine operations and third-party concerns. Search engines use
query and clickthrough data to target advertisements, to refine search
quality, and to personalize search. Prohibiting these uses outright could
have significant negative effects on users, including exacerbating search
engine bias. At the same time, third parties may have quite legitimate
interests in learning user identities, so that query-privacy policies reflect a
struggle between users and third parties for relative advantage. Given the
profoundly private information users entrust to search engines, balancing
these concerns will not be easy. A broader view of search makes clear the
competing values at stake.
Fourth, many legal theories raised by and against search engines turn
on the speech content of search recommendations. Search engines have
encouraged a view that such recommendations are subjective statements of
opinions about page quality and, as such, are entitled to substantial First
Amendment protection. But there is a sense in which it is precisely the
subjectivity of search rankings that make them problematic—the more
individually tweaked and the less automatic that results are, the greater the
concern that the search engine is using its privileged position to engage in
unfair discrimination against particular targets. The question of search
engine speech needs to be regrounded on a more stable foundation than
the subjective statement-of-opinion analysis alone can provide. A broader
view of search can provide a start.
Fifth, thinking about trademark disputes without placing them in
context is a recipe for trouble. Search engines provide enormous value to
consumers by literally reducing search costs. Trademark holders’ demands
for veto power over keyword sales must be understood as a tactic in the
ranking wars; giving them that power would frustrate the policies of
trademark law and hamper search innovation. At the same time, however,
simply excluding keyword sales and search engine manipulation from
trademark scrutiny altogether, as some courts have done, is also dangerous.
SEO tactics by providers increase consumer confusion and are socially
wasteful; it is easy to envision search engines and near relatives that flout the
goals of unfair-competition law. A broader view of search can point to a
healthier balance.
A. CLAIMS AGAINST SEARCH ENGINES AS FUNCTIONAL SUBSTITUTES
Multiple legal lines of communication exist between search engines and
other parties. Those concerned with one particular form of harm are not
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 53
limited to legal theories directly addressing that harm. If they can gain relief
against a search engine237 on another theory, it may be just as good.
Wherever in law this multiplicity appears, it raises a concern that parties not
be allowed to subvert one doctrine by appealing to another.
Consider first providers’ desire to prevent searching. Suits based on
unwanted access can often be interchanged with unfair-competition suits.
Many lawsuits by providers run together both forms of harm in their
complaints—the search engine both imposes technical burdens and
interferes with providers’ intellectual property.238 On the one hand,
competitive concerns motivated some of the most famous unauthorized-
access suits.239 On the other, providers who cannot use the law or self-help to
prevent access often turn to intellectual-property arguments. Thus, for
example, the Authors Guild has no ability to prevent Google from obtaining
physical access to books.240 Its lawsuit against Google is therefore exclusively
based in copyright infringement.241
A closer look at theories of unauthorized access shows this instability in
action. Pro-access advocates won a hard-fought victory when the California
Supreme Court in Intel v. Hamidi held that trespass to chattels would not lie
without proof of actual harm.242 Their victory is largely symbolic, given that
courts have been giving broad scope to anti-intrusion statutes and have been
willing to uphold browsewrap contracts.243 Hamidi notwithstanding, the
237. Or, in the engine’s case, gain a blanket immunity from suit.
238. See, e.g., Kelly v. Arriba Soft Corp., 336 F.3d 811, 817 (9th Cir. 2003) (raising copyright
theories predicated both on initial copying and on framing); Ticketmaster Corp. v.
Tickets.com, Inc., No. CV 99-7654 HLH(BQRX), 2000 WL 525390, at *2, *4 (C.D. Cal. Mar. 27,
2000) (raising copyright and trespass-to-chattels theories).
239. Register.com, Inc. v. Verio, Inc., 356 F.3d 393, 397 (2d Cir. 2004); eBay, Inc. v.
Bidder’s Edge, Inc., 100 F. Supp. 2d 1058, 1063 (N.D. Cal. 2000). The most promising unfair-
competition theories are not available to plaintiffs who provide factual information not
protectable by copyright. See Sw. Airlines Co. v. FareChase, Inc., 318 F. Supp. 2d 435, 440 (N.D.
Tex. 2004) (stating that “[f]are, route and scheduling information are all facts and thus not
copyrightable”); Ticketmaster Corp., 2000 WL 525390, at *2 (denying a copyright claim).
240. See 17 U.S.C. § 109(a) (2000) (authorizing the owner of a lawful copy the right to re-
sell that copy without the permission of the copyright holder). But see Sharon Billington, Relief
from Online Used Book Sales During New Book Launches, 29 COLUM. J.L. & ARTS 497, 497–98 (2006)
(arguing against empirical evidence and without considering the effects on libraries and critics
that publishers should be allowed to prevent resale of used books within six months of first
publication).
241. See Complaint at 2, Authors’ Guild v. Google, Inc., No. 05-CV-8136 (S.D.N.Y. Sept. 20,
2005), available at http://pub.bna.com/eclr/05cv8136comp.pdf. Cf. Ty, Inc., v. Publ’ns Int’l
Ltd., 292 F.3d 512, 515 (7th Cir. 2002) (considering a copyright suit against a collectors’ guide
to Beanie Babies containing allegedly infringing photographs thereof); Kevin Emerson Collins,
Cybertrespass and Trespass to Documents, 54 CLEV. St. L. REV. 41, 62–68 (2006) (discussing cases
involving the subsequent use of information obtained through improper access to physical
documents).
242. Intel Corp. v. Hamidi, 71 P.3d 296, 305–06 (Cal. 2003).
243. See supra Part II.B.1.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
54 93 IOWA LAW REVIEW [2007]
major search engines have accepted that a content provider who wants
search engines out can keep them out.244
Private third parties have also exploited overlapping doctrines;
copyright claims provide stronger relief against intermediaries like search
engines than do other private-law claims.245 Two ISP cases with search engine
angles show this phenomenon at work. In the pre-DMCA case of Religious
Technology Center v. Netcom On-Line Communications Services, Inc., the Church
of Scientology attempted to prevent the distribution of unpublished L. Ron
Hubbard manuscripts by asserting copyright claims against an ISP.246 The
Church has been a regular user of DMCA notices under Section 512 against
search engines.247 In Diebold v. Online Policy Group, an electronic-voting-
machine company attempted to use the Section 512 subpoena process to
suppress distribution of the source code to its voting machines.248 The
company relied on Section 512(d), the “information location tools”
provision applicable to search engines.249 Both plaintiffs were acting
principally to avoid the release of embarrassing secrets but were unlikely to
succeed on secrecy-based claims. Instead, copyright theories promised more
legal leverage.
Lawyers in search engine suits will not respect boundaries between legal
fields when framing their cases. Those who make law and policy for search
engines must be alert to these overlaps and end-runs. Considering the
various strands of search engine law together will help make such
possibilities clear.
B. THE PROS AND CONS OF DISCLOSURE AND MANDATED RESULTS
If we were certain we could identify instances in which search engines
returned incorrect results, the natural response would be to demand that
they return instead the results we knew to be correct. Even if we are not
certain we know which results are correct, we might know that some
particular results are incorrect and demand their deletion. Frank Pasquale
claims that subjects of a search deserve not to have search engines return
misleading information about them and identifies two cases in which the
correct search results can be specified precisely enough that he claims legal
intervention is warranted: searches on proper names and searches on
244. See supra notes 93–94, 108 and accompanying text.
245. See, e.g., 47 U.S.C. § 230(e)(2) (2000) (“Nothing in [the CDA intermediary immunity]
shall be construed to limit or expand any law pertaining to intellectual property.”).
246. Religious Tech. Ctr. v. Netcom On-Line Commc’ns Servs., Inc., 907 F. Supp. 1361,
1365 (N.D. Cal. 1995).
247. See Chilling Effects, Keyword:Scientologists, http://www.chillingeffects.org/dmca512/
keyword.cgi?KeywordID=10 (documenting a number of instances in which the Church used
such notices against search engines).
248. Diebold v. Online Policy Group, 337 F. Supp. 2d 1195, 1200 (N.D. Cal. 2004).
249. Id. at 1201.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 55
trademarked terms. Although he stops short of claiming that such searches
should instead return content provided by the person named or the holder
of the trademark, he does propose that they be allowed to annotate
misleading results with an asterisk.250
We can also make a related claim that doesn’t depend on having an
objective measure of “correct” and “incorrect” results. Start instead from the
observation that users depend on search engines to find information for
them. To know whether they should trust a search engine, they need to
answer the same question they asked the search engine to solve: What content
is available? Sometimes asking another search engine or having another user
ask the same search engine will suffice, but sometimes—as in personalized
search or when only one search engine indexes a particular kind of
content—those options aren’t available. In general, users need information
about a search engine’s inputs and its reasoning to make informed choices.
This claim leads, therefore, to an argument that search engines should
disclose detailed information about their algorithms.
Providers and search subjects also have legitimate reasons to want
greater disclosure. Google has been accused of manipulating search results
during litigation to make the judge at a crucial hearing unable to replicate
the behavior of which Google’s adversary complained.251 Providers who feel
they have been unfairly ranked for reasons unrelated to the quality of their
content may need access to operational details to evaluate whether they
really have been targeted. Third parties may want explanations as to why
unflattering content appears prominently. And many disputes about click
fraud cannot be evaluated without examining details of search engine
billing. Again, disclosure seems the natural remedy.
Natural, perhaps, but not necessarily a good idea. Excessive mandating
and disclosure can have dangerous consequences for the entire search
ecosystem. Search engines’ innovative interests mean that too much
disclosure can suppress their incentives to create new algorithms, reduce the
diversity of options, and limit users’ genuine choice among engines.
Mandated results are even more restrictive for competition and diversity, in
that they enforce uniform policies about results across engines. Moreover,
too much transparency in relationship to personalized search could lead to
the disclosure of user queries, raising privacy concerns.
250. Pasquale, supra note 3, at 117. But see Goldman, supra note 3, at 195–98. Pasquale’s
precise specification of these cases responds to Goldman’s critique that regulators cannot
identify correct results as well as search engines can by restricting intervention to clearly
identified mistakes.
251. See BATTELLE, supra note 1, at 184–86 (describing American Blinds’ allegations that
Google altered its search results to influence the judge’s ruling on Google’s motion to dismiss).
Search engine watchers know that there are many innocent potential explanations for such
behavior. Major search engines have multiple data centers, which may be out of sync with each
other. An update might have taken place between one query and another the next day. And so
on.
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
56 93 IOWA LAW REVIEW [2007]
Most serious of all are the consequences of mandated results and
disclosure for the SEO arms race. Search engine manipulators make their
living by reverse engineering search algorithms. Search engines are able to
preserve a layer of genuine, useful results through a combination of keeping
precise algorithmic details secret and changing their algorithms to foil
detected SEO techniques. Mandated disclosure undermines the former;
mandated results undermine the latter. Legal interventions here threaten to
hand search engine algorithms to manipulators on a platter. Even Pasquale’s
limited proposals are partially vulnerable to manipulation. What proof
would a search engine require of one’s real name before awarding an
asterisk? And what would stop manipulators from registering trademarks on
popular search terms on unlikely categories of goods? Consider a
registration of REAL ESTATE as used to sell lip gloss (a product category for
which it is fanciful, and thus registrable)—perhaps a pointless trademark but
excellent for search engine placement.
This is not the place to evaluate when disclosure and results mandates
are appropriate policy. For present purposes, it should suffice to note that
these remedies raise concerns that cut across many areas of search engine
law. There are reasons why they may be useful interventions and reasons why
they are dangerous. Considering one without considering the other would
be reckless.
C. USER PRIVACY CONCERNS IMPLICATE OTHERS’ INTERESTS
The privacy problems posed by private stockpiles of user data are well-
understood and do not require extensive rehashing here.252 What does
require attention is how the solutions to these problems may have different
inflections in the search engine context than in other domains. Users’
privacy interests must be understood in relation to other interests in query
data.
First, users’ own computers disclose to providers the users’ use of a
search engine and the query terms used. The “referer” information that a
browser by default gives to any site from which it requests a web page
includes the URL of the web page that referred the user to the provider’s
page. Most search engines include in the URL of the page displaying the
results the query terms that a user entered. This automatic leak of query
information—which can be blocked by technical measures either at the
search engine or at the user’s computer—means that search engines are not
the only institutions that can easily accumulate query data.253 Put another
252. See generally SIMSON GARFINKEL, DATABASE NATION (2000) (cataloging the threats to
privacy posed by database technology); DANIEL SOLOVE, THE DIGITAL PERSON (2004)
(examining the dangers of “digital dossiers” and proposing legal solutions).
253. Simply cutting and pasting a search result URL into one’s address bar will hide from
the provider the search query that led one to their site. One can also install software to much
the same end. See, e.g., Mozilla, RefControl, https://addons.mozilla.org/firefox/953/
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 57
way, the interaction of engine-to-user result flows with provider-to-user
content flows creates an additional user-to-provider query-data flow.
Second, third parties harmed by search may have legitimate interests in
learning some private information about users. This tension has been most
clearly noted in the case of flows of copyrighted information in the ISP
context; the Recording Industry Association of America (“RIAA”) has been
particularly active in attempting to learn users’ identities.254 Those whose
privacy is breached by a search also have an interest in learning about it:
they may need to take precautions against stalkers, they may need to take
action against the provider releasing this information, and they may need to
discourage searchers from searching.255 There is something uncomfortable
about a rule that assigns different weights to the privacy interests of search
users and search subjects.
Third, the uses to which search engines put their query-data warehouses
are relevant to users’ other interests in search. Massive stockpiles of queries
are useful for improving search. (Indeed, the AOL data release was neither
the product of poor security nor a concession to corporate pressure for
valuable data—it was an ill-advised attempt to further academic research
into better search technologies.)256 Extensive collection of query data is also
a prerequisite for personalizing search.257 Personalization of information
reception and its concomitant promotion of diversity, in turn, can be an
important technique for countering media bias.258 Thus, a privacy-mediated
(download page for RefControl, a program to hide referer [sic] information). Although the
blocking of such query leaks by search users is possible, it also seems to be quite rare.
254. See Sonia K. Katyal, The New Surveillance, 54 CASE W. RES. L. REV. 297, 297–98 (2003)
(discussing the RIAA’s attempts to locate infringing users through cell-phone records).
255. David Brin has written about this tension and has argued that the solution is greater
transparency in general. We cannot keep people from surveilling others but can at least let
those being surveilled know about it. See generally DAVID BRIN, THE TRANSPARENT SOCIETY
(1998).
256. See Katie Hafner, Tempting Data, Privacy Concerns; Researchers Yearn to Use AOL Logs, but
They Hesitate, N.Y. TIMES, Aug. 23, 2006, at C1 (discussing the academic ethics of research using
publicly released AOL query data).
257. Consider this in light of Amazon’s practice of recommending books by linking to
books that users searched for or bought in the same browsing session. On such “collaborative
filtering” systems, see generally JOHN RIEDL ET AL., WORD OF MOUSE: THE MARKETING POWER OF
COLLABORATIVE FILTERING (2002). Of course, ads can be personalized just as easily as organic
search results can. See, e.g., William Marra, Yahoo’s SmartAd Raises Privacy Concerns, ABC NEWS,
July 4, 2007, available at http://abcnews.go.com/Technology/Story?id=3342775 (describing
Yahoo’s search-advertising program, which allows advertisers to tailor ad placement based on
demographic information of users).
258. Cf. Yochai Benkler, Siren Songs and Amish Children: Autonomy, Information, and Law, 76
N.Y.U. L. REV. 23, 73–74 (2001) (arguing that personal autonomy and diversity are aided when
individuals can use a medium to pursue their own informational choices).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
58 93 IOWA LAW REVIEW [2007]
concern with preventing individual manipulation by search engines259 is in
tension with a concern with preventing manipulation by monolithic one-size-
fits-all information sources.
To repeat, problems of online privacy protection are subtle and
tangled. Considering the various threads of search engine law all at once
reveals just how many of them are connected to privacy in one way or
another. Once again, any rational attempt to make sensible policy (here,
privacy policy) in the search context demands careful engagement with
these many interests and pressures.
D. SEARCH ENGINE RESULTS AS SPEECH
Many legal questions involving search require a theory of search engine
speech. The First Amendment rights of search engines, users, and providers
may provide defenses to third parties’ attempts to impose liability for
harmful content flows. Search engines’ speech-facilitating roles may give
them a thumb on the scales in debates over access. And the conflicting
speech claims of search engines and providers will have a significant effect
on how we think about search engine rankings. Ultimately, a theory of
search engine speech will need to integrate all of these concerns.260
Such a theory is beyond the scope of this Article. Instead, this Section
considers the cross-cutting problems raised by one attempt at framing the
question. Google has asserted a theory of search rankings as subjective
statements of opinion. Under a claim of tortious interference with contract,
such as that raised in the Search King suit, it is a complete defense if the
allegedly harmful act consisted of protected speech. Statements of opinion
on matters of public concern are protected, unless provably false. Thus, for
example, a negative bond rating is not a statement that could be proven
“true” or “false” and thus cannot support defamation liability.261 Since a
search engine’s rankings are merely a claim about the engine’s subjective
assessment of pages’ relevance to particular users’ queries, goes the
reasoning, the search engine is not making a claim that could be shown
false—and is therefore protected.262
259. See Tal Zarsky, Online Privacy, Tailoring, and Persuasion, in PRIVACY AND TECHNOLOGIES
OF IDENTITY: A CROSS-DISCIPLINARY CONVERSATION (K. Strandburg & D. Stan Raicu eds., 2006),
available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=946498.
260. Substantial guidance may come from telecommunications law, which has long
struggled with intermediaries’ dual roles as speakers and as conduits. The greater interactivity
of search engines—whose rankings are driven by providers’ attempts to seek placement, by
users’ queries, and by search engine algorithms themselves—may require even more subtle
analyses.
261. Jefferson County Sch. Dist. No. R-1 v. Moody’s Investor’s Servs., Inc., 175 F.3d 848, 855
(10th Cir. 1999).
262. Search King, Inc. v. Google Tech., Inc., No. CIV-02-1457-M, 2003 U.S. Dist. LEXIS
27193, at *11–12 (W.D. Okla. May 27, 2003).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 59
The relationship of subjective opinion to objective fact, however, is not
simple. Thus, for example, Milkovich v. Lorain Journal Co.,263 while stating the
rule that the Constitution shields opinions, leaves in place two significant
exceptions. A statement of opinion may imply an underlying fact (the
Court’s example: “In my opinion John Jones is a liar.”264), and even a
statement of opinion may be false if not honestly held (the Court’s example:
“I think Jones lied,” where the speaker thought nothing of the sort265).
In this light, the Search King rule suggests several counterarguments.
First, as Search King alleged, the purported “opinion” is in fact the output of
a computer algorithm, and the computer is perfectly predictable and
objective. The court rightly dismissed this argument, distinguishing process
from result. The subjectivity entered the algorithm when it was
programmed, and although the intervening process is mechanical, what
emerges at the end are the subjective judgments made by Google
programmers about web page relevance and quality.266
Second, as Search King and KinderStart have alleged, search engines
themselves emphasize the objective quality of their results and should be
held to those statements.267 This argument has slightly more bite, given that
search engines have not been careful in their public statements.268 By
alleging claims of objectivity, provider plaintiffs also nudge their pleadings
closer to consumer-fraud causes of action in which the initial claim of
evenhanded objectivity is false, rather than the later ranking decision. Still,
this argument also fails. Search engines haven’t explicitly claimed that their
results are objectively correct and can craft their public-relations materials to
say that they don’t imply it, either.269
A more troubling counterargument, however, combines the first two. It
argues that the problem is hand manipulation of results.270 This theory
263. Milkovich v. Lorain Journal Co., 497 U.S. 1 (1990).
264. Id. at 18.
265. Id. at 20.
266. This view of the subjectivity of search results is consistent with an argument that they
involve sufficient selection and arrangement to satisfy copyright’s originality requirement.
267. See also Maughan v. Google Tech., Inc., 49 Cal. Rptr. 3d 861, 874 (Cal. Ct. App. 2006)
(noting Google’s emphasis that its search processes are completely automated).
268. Rebecca Tushnet has noted this tension. Tushnet, supra note 75. Raymond Nimmer
observes that Search King did not involve a claim by a user “who detrimentally relied on the
rankings themselves” and contrasts cases in which safety-ratings services “voluntarily assumed
[a] role [that] invited reliance by the public.” RAYMOND T. NIMMER, INFORMATION LAW § 10:77
(2006).
269. Moreover, this counterargument misses a basic reality of search engine business.
Search engines are trying hard to maximize the subjective satisfaction of users with their search
results, and on a query-by-query basis, it is extremely hard to find an objective “fact” in the
degree of user satisfaction.
270. See James Grimmelmann, Google Replies to Search King Lawsuit, LAWMEME (Jan. 9, 2004),
http://research.yale.edu/lawmeme/modules.php?name= News&file=article&sid=807, available
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
60 93 IOWA LAW REVIEW [2007]
abandons any claim to object to broad algorithmic decisions but argues that
specific deviations from algorithmic choices—or specific algorithmic tweaks
to hurt particular providers—should be actionable. The same programmers
both write the algorithms and tweak the results,271 so while there may not be
a clear line between the two, hand tweaks feel much slimier.
Why? Perhaps because the generic ranking algorithm provides a
baseline against which claims of later manipulation can be objectively
measured. In terms of the Milkovich analysis, the search engine is lying not
about the poorly ranked page’s quality, but about its own belief that the
page is of low quality. The ranking algorithm believes that the page is important, so
returning a worse rank is a lie about what the ranking algorithm believes. This way
of phrasing the argument captures its strong intuitive appeal. Even while
defending their rankings as subjective assessments, search engines have been
highly reluctant to make hand adjustments.272 There is something unsettling
about hand tweaks, whether or not that “something” rises to a level that
would permit a suit in tort.
Even this brief survey of one issue’s implications has stinted on other
significant connections. A concern with hand tweaks suggests that greater
transparency is a necessary tonic. The concern is also intimately related to
the ongoing struggles against SEO, since the most frequent algorithmic
changes are counters deliberately targeted at new SEO techniques. It
connects to fears of government censorship, to the individual deletions of
DMCA takedowns, and to broader questions of individual (and personal)
actions versus massive parallelism that arise in privacy and access to content.
A proper theory of search engine speech should consider these issues
together, rather than in isolation.
E. TRADEMARKS AND SEARCH ENGINES IN CONTEXT
The problem of searches on trademarked terms is, as noted above, one
of the most litigated issues in search. It also provides a case study in the
at http://web.archive.org/web/20040612081746/research.yale.edu/lawmeme/modules.php?
name= News&file=article&sid=807; Pasquale & Bracha, supra note 3.
271. See, e.g., William Slawsky, 20 Ways Search Engines May Rerank Search Results, SEO BY THE
SEA, Oct. 14, 2006, http://www.seobythesea.com/?p=334 (describing the systematic changes
that search engines might make to results after the basic relevancy algorithm but before
showing results to users).
272. See Google, An Explanation of Our Search Results, http://www.google.com/
explanation.html (stating that search results are generated completely objectively and are
independent of the beliefs and preferences of those who work at Google). Google drafted this
text to respond to public complaints that the first-listed result for the search “Jew” was an anti-
Semitic web site. This incident is significant in two ways. First, it revealed Google’s extreme
aversion to hand-tweaking its results, even though it would have been very easy to remove the
anti-Semitic web site from results or move it further down the list. Second, this annotation
required substantial public hand-wringing. (And note that Google now faces similar questions
about what standards it will use every time it must decide whether to add this annotation to a
results page.)
GRIMMELMANN_FINAL.DOC 11/21/2007 12:08 PM
THE STRUCTURE OF SEARCH ENGINE LAW 61
respective perils of too much and too little deference to search engines’
decisions.
Courts dealing with the trademark implications of provider and search
engine behavior have been much influenced by the Brookfield analogy of a
highway billboard. A user who is misdirected by an information-location tool
during a search for a trademarked term, the analogy asserts, is like a driver
who has been misdirected by a billboard to take the wrong exit. She may not
ultimately be confused about the source of the goods she acquires, but the
additional search costs of going back to find the true source outweigh her
desire for the “real McCoy.”273
This analogy misses two important features of search engines. First, it
misapprehends some of the technical realities of how various advertising
techniques actually translate into user visits. In particular, if keywords in
hidden metatags are billboards, they are invisible ones; most search engines
now ignore them.274 Similarly, contextual ads that are clearly disclosed as
such and do not use the trademarks in their text never appear to users as
billboards would; one might analogize them instead to billboards placed
near the plaintiff’s store but not using the plaintiff’s trademarks. Second, the
search costs involved in going back to find the originally desired source are
much lower online than offline—a few seconds of clicking rather than a few
minutes of driving—so far fewer users, even if diverted, will actually be
locked into the wrong source. Indeed, these advertisements reduce search
costs, first by offering users information about alternatives possibly
responsive to their queries and second by funding search itself. Contextual
ads are actually a substantial improvement, from a consumer-confusion
point of view, over earlier search business models such as direct results-
placement purchases and generic noncontextual banner ads.
Commentators and courts, however, have articulated a slightly
questionable basis for finding no liability. Instead of finding no consumer
confusion, they have found no use of the trademark, cutting off the
trademark inquiry at the threshold. The interactivity of the search engine’s
dealings with the user has created analytical confusion: any possible “use” of
the trademark is in the user’s search query, rather than in the results. It has
273. Brookfield Commc’ns, Inc., v. W. Coast Entm’t Corp., 174 F.3d 1036, 1064 (9th Cir.
1999).
274. Compare Pop Warner Little Scholars, Inc. v. N.H. Youth Football & Spirit Conf., No. 06-
cv-98-SM, 2006 WL 2591480, at *3 (D.N.H. Sept. 11, 2006) (“Google and other search tools
continue to associate defendants’ web site with plaintiffs’ marks [due to metatags] . . . .”), with
Eric Goldman, Outdated Metatags Don’t Infringe—Pop Warner v. N.H. Youth Football & Spirit
Conference, TECH. & MKTG. L. BLOG (Sept. 25, 2006), http://blog.ericgoldman.
org/archives/2006/09/outdated_metata.htm (criticizing Pop Warner and stating, “[S]earch
engines aren’t that inefficient or inaccurate given that THEY ARE SMART ENOUGH NOT TO
RECOGNIZE KEYWORD METATAGS IN THE FIRST PLACE”).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
62 93 IOWA LAW REVIEW [2007]
therefore seemed plausible to say that triggering ads or results based on a
trademarked query is not a “use in commerce” by the search engine.275
To see why this blanket rule may be inappropriate, consider a line of
interactive offline cases: trademark suits against restaurants that serve one
cola when a customer has requested another.276 The customer who receives a
Pepsi after ordering a Coke is a victim of passing-off; whether the deception
falls within the Lanham Act should not depend on whether the restaurant
has used the COCA-COLA trademark on its menu or whether the server
repeated “Coke” to confirm the customer’s order.
The search engine’s proper defense is that it is not misleading users,
not that it is not using the trademark. It is easy to imagine search engines
that deliberately cause serious confusion. Think of what would happen if
Froogle or Amazon—search engines specifically oriented toward finding
particular goods for purchase—were to adopt a policy of steering all
searches for COCA-COLA to purchase pages for substitute brands. The law
should not wholly ignore this possibility. Similarly, blatant SEO tactics cause
enormous consumer confusion—particularly when they push genuine
results entirely out of view—and a rule that such tactics are categorically
immune from trademark scrutiny because search engine spamming is not
trademark use seems perverse.
Finally, as in so many other contexts, the degree of concern increases
with the opacity of the search engine’s processes and the paucity of useful
alternative search engines available to users. Decisions that affect these other
matters affect the trademark inquiry. Thus, at the risk of sounding like a
broken record, I reiterate the theme of this Part: Looking at various strands
of search engine law together makes important connections clear.
Trademark law itself tries to incorporate many of these concerns, so
awareness of their practical effects improves the clarity of the doctrinal
trademark inquiry itself.
V. CONCLUSION
This Article has argued that search engine law is important and that it is
complicated. It is important because it exists at the point of convergence of
many strands of Internet law. It is complicated for the same reason. The
bulk of this Article has been an examination of the many doctrines from
which we must assemble a coherent law of search engines and of their many
interrelationships. Search engine law is a system with many moving parts but
275. See generally Stacey L. Dogan & Mark A. Lemley, Trademarks and Consumer Search Costs on
the Internet, 41 HOUS. L. REV. 777 (2004) (arguing that the “use in commerce” requirement
includes a requirement that a trademark be used as a trademark). But see Graeme B. Dinwoodie
& Mark D. Janis, Confusion Over Use: Contextualism in Trademark Law, 92 IOWA L. REV. 1597, 1609,
1629–36 (2007) (arguing that “use in commerce” incorporates no such requirement and that
such a requirement would be unwise, particularly for search engines).
276. See, e.g., Coca-Cola Co. v. Overland, Inc., 692 F.2d 1250, 1252 (9th Cir. 1982).
GRIMMELMANN_FINAL.DOC 11/20/2007 2:46 PM
THE STRUCTURE OF SEARCH ENGINE LAW 63
few degrees of freedom, and the challenge will be to satisfy as many
competing policy demands at once as possible.
Legal scholars have much work to do in the search space. Isolated
patches—the trademark law of keyword-advertising sales, access to computer
systems, intermediary liability for the flow of copyrighted materials, and a
few others—have received careful and sustained scholarly attention. But
these efforts must be connected and supplemented with equally thoughtful
analyses of the many other specific conflicts created by search. And they
must be connected with more overarching studies of larger themes in search
engine law.
This Article has repeatedly referred to some of these themes: the
tension between transparency and secrecy in search engine operations; the
relationship of competition among providers and among search engines; the
power of search engines to promote and infringe upon the privacy of users,
providers, and third parties; the role of search engines in enhancing and
inhibiting free speech; and the political economy of innovative freedom and
others’ claims upon search engines. A fuller discussion of these themes must
await other days and other articles. The need for such further study should
by now be apparent.
As of this writing, Google lists 15,800 results for “search engine law.”
That number will only increase.