Elliott v. Google: Trademarks | Outside IP Counsel

The Ninth Circuit held that verb use of google does not prove genericide because primary significance to consumers, not grammar, controls under the Lanham Act.

Elliott v. Google, Inc., 860 F.3d 1151 (9th Cir. 2017), decided May 16, 2017, answered a question that had hovered over one of the most valuable brands on earth: if everyone says “google it,” has GOOGLE become the generic name for searching the internet? The Ninth Circuit said no. Genericness is not a grammar test. A mark dies by genericide only when its primary significance to the relevant public becomes the name of a type of good or service, and the plaintiffs could not show that consumers understand “google” as the name for internet search engines as a class.

The decision is the modern framework for genericide litigation. It ties the doctrine tightly to the statutory primary significance test, separates linguistic habits from source perception, and explains why even a mark used daily as a verb by millions can remain a protectable, and famous, trademark.

At a glance

Case: Elliott v. Google, Inc., 860 F.3d 1151 (9th Cir. 2017)
Decided: May 16, 2017, amended June 14, 2017; opinion by Judge Richard C. Tallman, joined by Judge Paul J. Watford and by Chief District Judge Louis Guirola, Jr. (sitting by designation), with Judge Watford also filing a concurrence; affirming summary judgment from the District of Arizona
Holding: Verb use of “google” does not by itself constitute generic use; a genericide claim must show that the primary significance of the mark to the relevant public is a particular type of good or service (internet search engines), which the plaintiffs failed to do
Status: Final; the Supreme Court denied certiorari on October 16, 2017, Elliott v. Google, Inc., 138 S. Ct. 362 (2017)

A domain-name grab, a UDRP loss, and a cancellation suit

The dispute did not begin as a philosophical debate about language. Between February 29 and March 10, 2012, Chris Gillespie registered 763 domain names pairing “google” with other words and names, such as brands, people, and products (“googledisney.com” was one). Google objected, filed a complaint with the National Arbitration Forum under the Uniform Domain Name Dispute Resolution Policy, and won: the NAF agreed that the domains were confusingly similar to the GOOGLE trademark and had been registered in bad faith, and transferred them to Google on May 10, 2012.

David Elliott, later joined by Gillespie, responded by suing in the District of Arizona to cancel the GOOGLE registrations under 15 U.S.C. § 1064(3), which allows cancellation at any time if a registered mark “becomes the generic name for the goods or services.” Their theory was arresting in its simplicity: the public overwhelmingly uses “google” as a verb meaning to search the internet, verbs cannot identify source, and therefore the mark had become generic. The district court granted summary judgment to Google, and the Ninth Circuit affirmed.

The frame: genericness attaches to goods, not to acts

Judge Tallman’s opinion begins by restating the spectrum of distinctiveness from Abercrombie & Fitch Co. v. Hunting World, Inc., 537 F.2d 4 (2d Cir. 1976), and the peculiar career of generic terms. Some words are born generic; others, like aspirin in Bayer Co. v. United Drug Co., 272 F. 505 (S.D.N.Y. 1921), and the shredded wheat of Kellogg Co. v. National Biscuit Co., 305 U.S. 111 (1938), are victims of their own success, appropriated by the public as the name of the product itself. That process is genericide, and Congress fixed its measure in 1984 when it amended the Lanham Act to codify the primary significance test, repudiating the Ninth Circuit’s purchaser-motivation approach in Anti-Monopoly, Inc. v. General Mills Fun Group, Inc.

From the statute the panel drew two requirements that now structure every genericide case. First, the claim must be evaluated against a particular type of good or service. The plaintiffs framed their claim around an act, searching the internet, rather than a product category. But trademarks are registered for goods and services; the relevant question was whether “google” had become the generic name for internet search engines. Second, the challenger must show that the primary significance of the mark to the relevant public is that category. The Ninth Circuit shorthands this as the “who-are-you/what-are-you” test: a mark answers who the product comes from; a generic name answers what the product is. And because a registered mark enjoys a presumption of validity, the party seeking cancellation bears the burden of proving genericness by a preponderance of the evidence.

Verbs, grammar, and the dual-function mark

The plaintiffs’ central move was grammatical: an overwhelming majority of the public uses “google” as a verb, and, they argued, verb use is inherently non-source-identifying. The panel rejected the premise. Trademark law does not confine protection to adjectives. A term can serve a dual function, naming a product while at the same time indicating its source, and the statute asks about significance, not syntax.

The court then drew the distinction that gives the opinion its lasting analytical bite. An internet user might use the verb “google” in an indiscriminate sense, with no particular search engine in mind, or in a discriminate sense, thinking specifically of Google’s search engine. Only the indiscriminate kind even arguably speaks to genericness, and even then it addresses the wrong question unless it shows what the public understands the word to mean as applied to search engines. Evidence that most people say “google it,” standing alone, cannot carry a genericide claim, because a speaker can verb the brand while knowing perfectly well whose product it is.

Measured against that framework, the plaintiffs’ record collapsed. Their expert’s survey showed that more than half of respondents would use “google” as a verb to describe searching the internet, but verb use did not reveal whether respondents thought of the word as a brand or a category. Google’s own consumer survey, by contrast, showed that over 90 percent of respondents recognized GOOGLE as a brand name rather than a common name for search engines. The plaintiffs’ dictionaries listed “google” as a verb but tied the definitions to the Google search engine or flagged the trademark. Their media examples and linguists’ declarations again proved verb use, not primary significance. And their argument that no efficient substitute exists for the verb missed the mark: competitors do not need to call their products “googles,” and consumers have no trouble naming Bing or Yahoo as search engines. There was, the court concluded, no genuine dispute for trial.

Judge Watford concurred with one caveat. He would not foreclose the possibility that evidence of indiscriminate verb use could, in some future case, tell a jury something about primary significance; he simply found the plaintiffs’ evidence too insubstantial to defeat summary judgment given the lopsided survey results and their own experts’ concessions. The Supreme Court denied certiorari in October 2017, leaving the framework intact.

What Elliott settled in genericide doctrine

Elliott is now the standard citation for three propositions. Genericide claims must be tethered to the registered goods or services, not to a colloquial activity. Grammatical drift, however widespread, is not a proxy for genericness; the statutory question is always primary significance to the relevant public. And survey evidence sits at the center of that inquiry: a Teflon-style brand-recognition survey showing strong source association will ordinarily overwhelm usage evidence that never asks what consumers think the word names. The Supreme Court’s later decision in United States Patent and Trademark Office v. Booking.com B.V., 591 U.S. 549 (2020), reinforced the same consumer-perception premise from the registration side: whether a term is generic turns on its meaning to consumers, not on per se rules.

Open questions

The concurrence marks the live seam: whether, and how, indiscriminate verb or noun use could ever be packaged into admissible proof of primary significance. A challenger with a rigorous survey isolating indiscriminate use might present the question the majority did not need to decide. The opinion also leaves unresolved how the framework applies to marks whose product category is genuinely contested, where defining the genus (search engines? general web search? answer engines?) can predetermine the outcome; that definitional fight has only sharpened as AI assistants blur what counts as “searching.” And Elliott offers no comfort on the outer bound of dilution-by-success: it tells owners of famous marks that verbing alone will not kill them, but it does not say how much indiscriminate usage, accumulated over decades, would.

Implications for brands and businesses

Police the record, not the grammar police’s approval. Verb use alone will not doom a mark, but indiscriminate use remains ammunition. Style guides, “Google it on Google Search” phrasing, trademark notices, and corrections to publications all build the evidentiary file that wins the survey war later.
Genericide fights are won with surveys. A challenger must prove what the term primarily signifies for the registered goods; an owner should be ready with Teflon-format brand-recognition evidence. Google’s roughly 90-plus percent brand recognition made summary judgment possible.
Frame the genus before your opponent does. Because genericness is measured against a particular type of good or service, the definition of the category is often outcome-determinative. Registrants should keep identifications precise, and challengers cannot substitute an “act” for the goods.
Fame is not fragility. Elliott rejects the intuition that ubiquity converts a brand into language. But the safest course for a market-defining product is to give consumers a generic noun to use (search engine, ride-hailing app) so the mark never has to serve as the category’s name.

Frequently asked questions

Did the court say using a trademark as a verb is safe for the brand owner? No. It held only that verb use does not automatically prove genericness. Widespread indiscriminate verb use, where speakers mean any product in the category rather than the brand, remains evidence a challenger could marshal. That is why brand owners still police verb and noun uses in style guides and enforcement letters: they are managing the evidentiary record against a future genericide claim.

What is the who-are-you/what-are-you test? It is shorthand for the primary significance inquiry codified at 15 U.S.C. § 1064(3). If consumers understand a term as answering who a product comes from, it functions as a mark; if it answers what the product is, it is generic. A mark is cancelled as generic only when its primary significance to the relevant public is the name of the type of good or service rather than an indicator of source.

How does Elliott square with cases like aspirin and escalator becoming generic? Those terms crossed the line because the public came to use them as the name of the product category with no source in mind, and often no alternative generic name existed. Elliott confirms the same standard but shows the proof must target the category: the plaintiffs never showed that consumers think google names all search engines. Google’s survey evidence showed the overwhelming majority still recognize it as a brand.