GEICO v. Google Opinion (Finally) Issued

Government Employees Insurance Co. v. Google, Inc., No. 1:04cv507 (E.D. Va. Aug. 8, 2005).


On December 15, 2004, Judge Brinkema made an oral ruling in this case and promised to produce a written opinion in “a brief amount of time, which given the holiday season might be two or three weeks, although we’re going to try to do it sooner than that.” Eight months later, we finally got the written opinion. Why so long?

Was it worth the wait? In a word, no. The opinion does very little to clarify or add to the court’s oral ruling from December. So we really don’t learn much from this opinion, and as a result, I can’t see why this opinion will affect the parties’ settlement negotiations.

What the Court Said

In an August 2004 opinion, Google lost its effort to dismiss the case because it does not use GEICO’s trademarks in commerce. Therefore, this opinion only addresses the likelihood of confusion standards.

In determining the applicable standard, the court does some arm-waving. It appears to endorse GEICO’s position (from Brookfield and other cases) that an “initial interest confusion” case moots the need to use the standard likelihood of confusion multi-factor test. However, the court merely describes GEICO’s position without directly endorsing it.

Instead, after describing that position, the court summarizes by saying: “To prove likelihood or absence of confusion, initial or otherwise, parties commonly introduce the results of customer or potential customer surveys.”

It’s true that surveys are a standard method of proving likelihood of consumer confusion, and I’m glad the court required some empirical proof rather than mere intuition-driven arguments that pervade most initial interest confusion cases. However, what legal standard is the court using? If it is bypassing the multi-factor likelihood of confusion test, what test is it using? The court doesn’t say.

One other odd thing about this statement. What’s this about proving the absence of confusion? The burden is on the plaintiff, not the defense. Why would anyone need to prove the absence of confusion?

Having punted on describing the applicable standard except to say that surveys are important, the court spends almost all of the rest of the opinion analyzing the plaintiff’s survey. The survey presented consumers with a screen shot showing the results from a search for “GEICO.” 67.6% said they expected to reach GEICO if they clicked on the sponsored links, and 69.5% thought that the sponsored links were either for GEICO or affiliated with GEICO. Further, 20.1% said that to purchase insurance from GEICO, they would click first on one of the sponsored links.

However, the court then recounts the various reasons why the survey is not credible, including the poor construction of a control group, “demand effects” (when the survey suggests desired answers) and “order effects” (when answering a question affects subsequent Q&A), and distortions in the appearance of the screen shot.

All told the court has “serious doubts about the accuracy of the survey results’ reflection of actual users’ experiences with and reactions to the Sponsored Links.” As a result of the survey’s weaknesses, the court says that the plaintiff failed to establish likelihood of confusion about keyword triggering.

Indeed, in FN 15, the court notes that none of the control group was confused when Nike ads were displayed in response to a GEICO keyword. The court says, “This absence of confusion when faced with NIKE ads alongside organic listings for “GEICO” also refutes the allegation that the use of the trademark as a keyword, without more, causes a likelihood of confusion.”

FN 15 is interesting because it suggests that the court needs to look at the ad copy to make a likelihood of confusion determination. I completely agree with this, but I would go one step further–each individual ad must be reviewed for confusion. This should mean that the court can’t make any broad sweeping pronouncements about confusion based on the inclusion of a trademark in the ad copy; it should depend on consumer responses to specific ad copy.

The court doesn’t embrace that level of nuance. Instead, the court continues, “Despite the many flaws in its design, the survey’s results were sufficient to establish a likelihood of confusion regarding those Sponsored Links in which the trademark GEICO appears either in the heading or text of the ad.”

This is a non-sequitur. There’s nothing in the court’s opinion or the discussion of the survey that supports this finding. Unfortunately, the court does nothing to explain its thinking about this conclusion or how the survey supports it. The only thing that the court notes is “the extremely high percentages of respondents who experienced some degree of confusion when viewing [ads displaying GEICO in the text].” Further, Google did not introduce any rebutting evidence on that point.

As a result, the court rules for GEICO that “Google may be liable for trademark infringement for the time period before it began blocking such usage or for such ads that have slipped or continue to slip through Google’s system for blocking the appearance of GEICO’s mark in Sponsored Links.”

Although the court has concluded that the ads create the requisite level of consumer confusion, Google has not lost this case. Open questions include whether Google is liable for its advertisers’ use of GEICO’s trademarks (i.e., if Google is contributorily liable for the confusion), the number of ads at issue (given some changes in Google’s practices), and the remedies (including, I think, if Google can claim to be covered by the printer/publisher defense that would eliminate damages altogether and leave open only an injunction).


I’m sure both Google and GEICO are frustrated by this opinion, because it does very little to clarify the court’s ruling from December. Indeed, although it provides support for its ruling, the opinion adds almost nothing new in terms of insights into the judge’s view.

What I find interesting is that the court, like the 1-800 Contacts Second Circuit panel, knows that this is an important case and yet goes out of her way to avoid providing any broader principles to help future parties. She writes “the Court emphasizes that its ruling applies only to the specific facts of this case, which include the unique business model employed by plaintiff and the specific design of defendant’s advertising program and search results pages.”

This may be a healthy reaction to the 1990s cases where judges, knowing very little about the technology, all too often wrote broad but ill-conceived opinions that took years to correct. Yet, we depend on our court system to help us understand the law, and narrowly-conceived and -written opinions really don’t give us those widely-felt benefits.

In any case, I’ll characterize this opinion as a win for Google for three reasons: (a) the opinion isn’t much worse for Google than the oral ruling, (b) Google succeeded in knocking out the most significant part of the case in the oral ruling, and (c) Google still has some defenses that may allow it to win outright in further proceedings.

GEICO’s press release, which GEICO uses for saber-rattling: “GEICO will continue to aggressively enforce its trademark rights against purchasers of its trademarks on search engines and against search engines that sell GEICO’s trademarks to advertisers…We continue to believe that the sale of GEICO’s trademarks to its competitors is wrong and a violation of federal and state law and look forward to litigating that issue in future cases.”

UPDATE: Google has posted on its blog an explanation of why it thinks this ruling is a win for Google and how some of the (very few) popular news stories have mischaracterized the ruling as a loss. (Thanks to Cedric for catching this).