by Paul Alan Levy
In its decision issued last week about the application of trademark law to the operation of online search engines, Multi Time Machine v. Amazon the Ninth Circuit has taken a dangerous step backwards. The case might, however, prove to be the vehicle by which that court wipes out the bad precedent on which the case is founded and thus protects consumers from a range of abusive uses of trademark law to suppress both consumer commentary and effective competition.
The Majority and Minority Opinions
Multi Time Machine v. Amazon arose out of searches done by potential customers of Amazon’s vast commercial web site using “mtm special ops,” the name of a fancy watch made by the plaintiff, Multi Time Machine, as a search term. Multi Time Machine apparently does not allow its watches to be sold through Amazon. Applying the Ninth Circuit created doctrine of initial interest confusion, and performing a results-oriented recitation of the Sleekcraft factors — that circuit’s version of the “likelihood-of-confusion factors” — the panel majority, in an opinion by Judge Bea, ruled that there was sufficient evidence to preclude summary judgment on the question whether there was likely confusion on the part users seeing this list of search results about whether one or more of the search results (the panel did not say which one was confusing) would lead to a page where they could buy or learn more about plaintiff’s product. Although each of the eight listings for non-MTM watches identifies the MTM competitor’s brand of the watch that could be accessed through the hyperlink included in the listing (two of the links are for books rather than watches), the majority singled out the fact that the search term appears three times on the page, and the absence of any disclaimer provided by Amazon specifying that none of the ten main links leads to a page where plaintiff’s product can be bought.
Although there was no evidence of actual confusion about the search listing page itself, nor any evidence that anyone visiting the linked pages bought a competing watch under the misimpression that it was an MTM watch, the majority relied on the doctrine of initial interest confusion, according to which the fact that a search engine user’s interest in a competitor’s web page might have been piqued by a search result produced by use if the mark as a search term can be a sufficient basis for finding Lanham Act liability. And the search results listings might be actionably confusing, as the majority saw it, in that some users might have wondered whether, because these results were produced in response to use of the mark as a search term, MTM might have been acquired by the companies making these watched.
The majority opinion includes several assertions about how Internet users interact with search engines and what they might have thought. Yet the opinion contains no citations to or references to the record apart from an expert report that is referenced only vaguely; hence it is not clear what the basis was for these assertions about what search engine users might think.
The majority opinion drew a strenuous dissent from Judge Silverman, who reprinted the Amazon search results page that both opinions discuss as part of his opinion (page 24). For Judge Silverman, it was enough that each of the search results for watches lists the name of the company making them—a judge looking at that page, he said, could be sufficiently confident that it did not create a likelihood of confusion that summary judgment was warranted on that issue. He argues against applying the Sleekcraft factors because that test was adopted for cases in which the question is whether the marks of competing brands are similar enough to cause likely consumer confusion about source. Judge Silverman refers repeatedly to a a pair Ninth Circuit decision from 2010 and 2011, Toyota v. Tubari and Network Automation v. Advance Systems Concepts, which had found no likely confusion in situations involving search engines, in significant part by assuming that Internet users are sophisticated about online commerce and not easily confused. (The majority distinguished Network Automation on the ground that it involved a preliminary injunction on which the plaintiff had to prove likelihood of success, while MTM had only to show that there were sufficient genuine issues about a material fact to defeat summary judgment; it made no mention of the fact that Toyota v Tubari was an appeal from a judgment after a bench trial.) Judge Silverman taunts the majority for its speculation about consumers believing that MTM was connected to one of the competing companies because the record contained no evidence of such confusion. He also notes that the list of search results included two books, a plain signal to the sensible search engine user that the listings might not be MTM watches.
Analysis of the Decision
As I see it, the dissent reached the right result but not entirely for the right reasons—its criticism of the majority opinion was not nearly searching enough. The problem goes back to a Ninth Circuit decision from 1999, that court’s first major foray into the interplay between trademark law and the various devices for locating web sites of interest. In Brookfield Communications v. West Coast Video, the parties were contesting which of them had priority for use of the trademark MovieBuff on their respective products, and the parties had briefed only that issue. But in addition to addressing that question, the panel chose to focus its attention on how internet users would react to seeing the trademark in the junior user’s domain name and to reaching its web site through a search engine as a result of the inclusion of the trademark in its meta tags, issues that the parties had ignored in their briefs. The panel opinion drew from some early decisions in other circuits about companies that had deliberately misidentified themselves in sales calls by using another company’s trademarked names in order to get their foot in the door, while at the same time articulating a series of factual assumptions about Internet users that had no basis either in the record or in consumer research, and created the doctrine of “initial interest confusion” which permitted as finding of trademark infringement when possible confusion that brings an Internet user to a web site even if the user experiences no confusion when making a purchase from a competitor (or, indeed, even if the user makes no purchase at all).
Based perhaps on the unfamiliarity of the judges themselves with the operation of the Internet, and based perhaps with some justification on the newness of the Internet in 1998 and the still-relatively-rudimentary character of finding devices such as search engines whereby Internet users could locate web sites of interest, the Brookfield Communications panel proclaimed that Internet users are both fickle and unsophisticated. The majority was ready to assume that any Internet user who enters a trademark into a search engine is necessarily looking for the web site operated by the trademark owner; that if the search engine results display other sites in addition to or instead of the markholders’ site, these other search results represent potential “diversions” from what the user must really have been looking for; and that if a competitor’s site appears in the search results, and is chosen by the user, then the competitor’s efforts to secure that user selection represents a firm of unfair trickery against which trademark law should protect. The opinion assumed further that, because Internet users are fickle and unsophisticated, if they clicked through to a web site that was not one that they really wanted, they would either give up and stop searching for the trademark holder’s web site, or even make a purchase at a competitor’s site. The opinion drew an analogy between misleading search engine results and a misleading highway sign suggesting that a given business could be visited by using a particular highway exit, when, in fact, it was only a competitor that could be found at that highway access. The majority assumed that the harm incurred as a result of a misleading Internet link is comparable to the harm of the highway sign. In all of this analysis, there was not a single citation to the record or to any empirical analysis supporting the majority’s assumptions.
The Brookfield panel seized on the concept of “initial interest confusion” to address this problem, holding that the non-authorized user of a trademark can be held liable for trademark infringement based on a likelihood of initial interest confusion alone, even if the user never purchases a competitor’s goods and even if they purchase competitor goods despite understanding that it is a good from the competitor and not from the trademark user. For several years, the Ninth Circuit and several other courts applied and even expanded this analysis, building up a considerable body of precedent that extended trademark infringement to a variety of circumstances in which Internet users were protected against being exposed to competing products and even to criticisms of the trademark holder in the name of preventing initial confusion.
Critiques of Initial Interest Confusion
There is a rich vein of scholarly literature, including leading articles
by Eric Goldman, Jennifer Rothman, and by Mark Lemley co-authored with Mark McKenna and with Stacy Dogan, addressing the flawed assumptions of Brookfield and its progeny, and questioning, indeed, whether that doctrine has been applied consistently or whether it simply serves as a buzzword that gets courts to the destinations they had chosen in each case. At the same time, several courts outside the Ninth Circuit began to question the factual assumptions underlying initial interest confusion. Courts recognized that the real-life consequences of landing at the “wrong” website as a result of initial interest confusion were trivial. Even search engine users who have definitive objectives about the web sites that they hope to locate by conducting searches using search engines recognize that the engine will return a range of results, not only on the first page of listings but after that as well, and that not every search engine result will be the page they were hoping to reach; the listing contains those items from the search engine’s database that the search engines ranking algorithm designates as being most relevant to the search terms. It is even possible that none of the listings in the page of search results will be page that the user wants. And if they follow links from the list of results that take them to web pages that are not what they want, they will go back to the search engine, perhaps by using the “back” button on the browser. And if a given search engine consistently fails to give users what they want, they will perhaps change search engines.
How the Doctrine of Initial Interest Confusion Can Hurt Consumers
From a consumer perspective, elevating the doctrine of initial interest confusion to a freestanding basis for finding trademark infringement can be counterproductive and even harmful. It gives the owners of trademarks the ability to seek judicial rulings that keep information about competing products, and even criticisms of the trademark holder and its products, out of search engine listings where consumers can find them. However, such information might be just as useful to search engine users as the information found on the trademark owner's own web site; it might even be what the search engine users hoped to find. Even if we hypothesize that the only reason why some people use a trademark as a search term is that they hope to find the trademark user's product and buy it, other users might recognize the trademarked name as an example of a product that they want but be perfectly content to find and buy competing products of the same type that are better, or cheaper, or more readily accessible. Still other users might want to obtain information about the trademark holder so that they can assess its desirability, or find others who are unhappy about the product and make common cause with them. When we first starting litigating trademark cases brought against Internet gripe sites, for example, trademark holders consistently invoked the concept of initial interest confusion as a basis for preventing gripe sites from coming to the attention of Internet users, thus making it harder for Interney users to obtain negative information about the brand and its owner.
Trademark law finds its social justification as a protection for consumers against being misled about the identity of purveyors of commercial information; consequently, trademark law needs to accommodate the many different reasons for using search engines. Accommodating careless Internet users who don’t pay attention to small differences, and who are too lazy to use the “back” button on their browsers when they reach the wrong location, can mean suppressing access to useful information that other search engine users want. And if trademark law is subverted to that end, then trademark law is not serving the interests of consumers at large; it is accommodating a few (and accommodating trademark owners who want to suppress information) but hurting the many. And considering how trivial the consequences of initial interest confusion are (even if its occurrence had empirical support), it is hard to see why trademark law should suppress access to useful information when the harm to consumers is so small.
Indeed, even the use of initial interest confusion to defeat summary judgment can have the effect of suppressing fair competition and fair criticism. Trademark litigation is notoriously expensive – the biannual surveys conducted by the American Intellectual Property Law Association tell us that the average cost of taking a trademark case to trial is in the mid-six figures. I don’t feel much solicitude for defendants like Amazon, which can easily afford to defend themselves, but when an Internet critic, or a small or medium sized business, is faced with having to go to trial on a trademark claim, the impact can be ruinous. The mere threat of bringing such litigation by invoking the doctrine of initial interest confusion thus enables trademark holders to suppress access to criticism or information about competing brands.
Initial Interest Confusion Loses Judicial Favor
As a result of these considerations, and the issuance of trial court rulings undercutting the empirical foundations of the initial interest confusion concept, some courts of appeals expressed their doubts about the doctrine of initial interest confusion itself. These doubts also began to seep into Ninth Circuit rulings as well, which articulated factual characterizations about Internet users that contradicted Brookfield. For example, in Interstellar Starship v. ePix and Entrepreneur Media v. Smith, Ninth Circuit panels agreed that Internet users are sophisticated about how the Internet works and that they notice and appreciate the significance of small differences among names and listings. In a concurring opinion in Playboy Enterprises v. Netscape Communications, Circuit Judge Marsha Berzon wrote a scathing criticism of Brookfield Communications, questioning both the highway sign analogy and and ensuing expansive applications of initial interest confusion This thinking culminated in the pair of decisions by Judges Kozinski and Wardlaw, cited by Judge Silverman’s dissent, Toyota v. Tubari and Network Automation; the latter opinion included this critique:
in the age of FIOS, cable modems, DSL and T1 lines, reasonable, prudent and experienced internet consumers are accustomed to such exploration by trial and error. They skip from site to site, ready to hit the back button whenever they're not satisfied with a site's contents. They fully expect to find some sites that a
ren't what they imagine based on a glance at the domain name or search engine summary. Outside the special case of … domain [name]s that actively claim affiliation with the trademark holder, consumers don't form any firm expectations about the sponsorship of a website until they've seen the landing page — if then.
It is difficult to understand how the majority opinion in Multi Time Machine is consistent with this analysis; the distinction between the procedural posture of the two cases hardly explains the difference in approach.
Applying a Realistic View of Search Engine Use to the Facts of Multi Time Machine v. Amazon
In writing this final section of this post, I emphasize that I have not yet had the opportunity to study the appellate record; I have not read the appellate briefs or the record excerpts, not to speak of reading the summary judgment papers below and all of their exhibits. My comments are limited to the facts described in the majority and dissenting opinions (generally without citation to the record), and particularly the list of search results reproduced in Judge Silverman’s opinion. The majority opinion cites an “expert report” but does not describe it with sufficient detail to allow any conclusions about what it signifies — the expert appears to have concluded that the search results listing is misleading but as described by the majority this appears to have been the expert’s own opinion and not a reflection of a consumer survey. (A footnote near the beginning of the majority opinion cites some search results obtained shortly before the opinion was issued, saying that they are subject to judicial notice. The quoted language in these search results sounds affirmatively misleading).
From this information, it appears that what users of Amazon’s internal search engine see when they enter the term “mtm special ops” as a search term is a listing of eight watches, each designated with a competing brand name, and two books, also listed with its publishers’ information. Each search result has a hyperlink that takes the user to the page on Amazon’s own web site where the item can be purchased. And at the very bottom of the page is a “sponsored listing” for “Tactical Watches by MTM” that appears to lead Internet users to one of the plaintiff’s own web pages.
As between the two opinions, it strikes me that Judge Silverman has the better argument about how Internet viewers would likely perceive this page. The very presence of two books in the search listing is a signal to the user that not every item in the listing is one of the plaintiff’s products. Moreover, each of the competing products is labeled with the competing brand name.
The majority makes much of the absence of a specific disclaimer stating that none of the watches linked on the page is one of the plaintiff’s products. Given those labels, it is hard to see why the specific search listings themselves should contain disclaimers of affiliation with the plaintiff; and besides disclaimers are generally useful in avoiding confusion about improperly labeled products; the absence of a disclaimer is generally not a basis for finding likelihood of confusion. Perhaps what the majority wanted to see was a disclaimer stating that none of the listings on that page reflected one of plaintiff’s products, but that disclaimer would have been inaccurate given the fact that there is a listing for plaintiff’s products as a “sponsored link” at the bottom of the page.
But even more important, even if Amazon's algorithms could have framed the page could have been framed better, why should trademark law require it to do more? Purchases cannot be made from the results listing; when Amazon's users visit the pages linked from the search results, they will see clearly that the watches were made by Casio, Luminox and MTM's various other competitors. And if what they really wanted to find MTM watches only, they will either go back to the search page and click on the sponsored link, or they will use a different search engine, such as Google.
The majority opinion is silent about its implications for other search engines. Will they have to post disclaimers about the inadequacy of search listings because they might not include links for trademark holders' web sites that might be what is of interest to the search engine users?
Will The Ninth Circuit Finally Rid Us of Brookfield?
In the end, this is a decision that rests squarely on the concept of initial interest confusion, invoking outdated understandings of how Internet users see search engine listings and contradicting recent Ninth Circuit decisions. Although Eric Goldman's first published column about the case suggests that Amazon can limit its exposure to such cases in the future by adding the specific disclaimers that the majority demands, we can hope that Amazon will seek en banc review; from what I have seen so far, this case could present a good vehicle for the Ninth Circuit to decide whether to retain the doctrine.
The Ninth Circuit created this quandary with its Brookfield Commuications decision. That court should undo the serious harm to consumer interests that this decision fostered. Indeed, over the past sixteen years, various panels have treated the doctrine in various way; although as of 2011 it appeared that initial interest confusion was close to being a dead letter in the Ninth Circuit, the doctrine is now very much alive again, and it appears that the outcome of appeals in which the doctrine figures may depend on which three judges serve on the panel. Given the frequency of trademark litigation in the Ninth Circuit, that approach is untenable. En banc review is needed to sort the issue out.