Keyword Advertising Disputes

The two questions, and which one is still live

Keyword advertising litigation divides cleanly. First: is buying or selling a competitor's mark as a keyword a "use in commerce" at all? Second: is the search results page that follows likely to confuse anyone? The first question is largely settled against defendants. The second is now settled overwhelmingly in favor of them.

That asymmetry drives how these matters are worked up. A plaintiff who pleads that the defendant bought its brand term has pleaded almost nothing, because the purchase is not the injury. The injury, if there is one, lies in what appeared on the screen and what happened next. The evidence that decides the case is therefore a paired record: a reconstruction of the advertisement as displayed — headline, description, the "Ad" label, the displayed URL, the surrounding results, the landing page — and a quantification of how many people saw it, clicked it, and did something afterward.

Use in commerce, and why it stopped being a defense

Rescuecom Corp. v. Google Inc., 562 F.3d 123 (2d Cir. 2009) came up on appeal from a Rule 12(b)(6) dismissal. The Second Circuit held that Rescuecom had adequately pleaded use in commerce because Google sold and recommended the mark as a keyword through its suggestion tool — Google "uses and sells Rescuecom's mark 'in the sale [of Google's advertising] services rendered in commerce.'"

What the court expressly declined to decide is the part practitioners forget: "We have no idea whether Rescuecom can prove that Google's use of Rescuecom's trademark in its AdWords program causes likelihood of confusion."

The opinion's appendix traces the 1962 statutory rearrangement and concludes that the restrictive definition of "use in commerce" governs qualification for registration rather than infringement liability. That is why the defense is not a serious one in the Second Circuit. It has not gone entirely quiet: a 2024 Ninth Circuit concurrence argued the court should reconsider whether keyword bidding is use at all, since the purchaser never displays or affixes the mark — Google does.

The factors that decide a keyword case

Network Automation, Inc. v. Advanced Systems Concepts, Inc., 638 F.3d 1137 (9th Cir. 2011) is the pivot: the defendant bought a competitor's mark as a keyword to advertise its own software, and the Ninth Circuit reversed a preliminary injunction.

The court rejected the three-factor "Internet troika" inherited from Brookfield, holding that "it makes no sense to prioritize the same three factors for every type of potential online commercial activity." It identified the four factors most probative in keyword cases — strength of the mark, evidence of actual confusion, type of goods and degree of purchaser care, and the labeling and appearance of the advertisements and the surrounding context on the screen.

Two of those four are, in practice, technical questions. On purchaser care the court declined to assume internet users exercise minimal care, noting "there are many contexts in which it no longer holds true," and pointed to the nature and cost of the goods. On labeling it held that the appearance of an advertisement "includes more than the text of the advertisement, and must be considered as a whole." That is an evidentiary assignment: reconstructing the whole page as it rendered, on desktop and on mobile, which differ materially.

Three separate proceedings with the same party name

Counsel conflate these constantly, which produces briefs citing an antitrust holding for a trademark proposition. Keep them apart.

The trademark case. 1-800 Contacts, Inc. v. Lens.com, Inc., 722 F.3d 1229 (10th Cir. 2013). Keyword purchase alone did not infringe; the court relied on evidence that the ads diverted approximately 1.5 percent of users. On contributory infringement it reversed as to affiliates whose ad copy did display the mark, holding that "a rational juror could find that Lens.com failed to take appropriate corrective action when notified of the issue." What the advertiser did about known misuse by affiliates mattered more than what it bought.

The FTC administrative proceeding. Matter 141 0200, Docket No. 9372. The complaint, filed 8 August 2016, charged bidding agreements with at least fourteen competing retailers that eliminated competition in search auctions. An ALJ initial decision issued 20 October 2017 and the Commission's opinion and final order followed on 14 November 2018. The mechanism was pairwise settlements in which each party agreed not to bid on the other's marks and to employ negative keywords — instructions telling the ad system not to serve on specified queries.

The appeal. 1-800 Contacts, Inc. v. FTC, 1 F.4th 102 (2d Cir. 11 June 2021). The Second Circuit vacated the order, holding that the Commission improperly treated the agreements as inherently suspect. Trademarks, it reasoned, "are by their nature non-exclusionary," and agreements protecting trademark interests are "common, and favored, under the law," requiring full rule-of-reason analysis.

Where the law landed in 2024 and 2025

Two decisions in October 2024 largely closed the field, and the Supreme Court declined to reopen it.

In 1-800 Contacts, Inc. v. JAND, Inc., No. 22-1634 (2d Cir. 8 October 2024), the Second Circuit affirmed dismissal at the pleading stage. It held that "the mere act of purchasing a competitor's trademarks in the context of keyword search advertising does not constitute trademark infringement," and that similarity of marks "should be assessed as it relates to the paid advertisement's appearance on the result page," not to the purchased keyword. The defendant's ad carried the "Ad" label and displayed its own domain, and the landing page displayed its own mark; dissimilarity was dispositive despite allegations of mark strength and bad faith.

In Lerner & Rowe PC v. Brown Engstrand & Shely LLC, 119 F.4th 711 (9th Cir. 22 October 2024), the Ninth Circuit affirmed summary judgment for a law firm that bid on a competitor's name without using the mark in ad copy. The record numbers are the most useful thing in the opinion for anyone building or attacking a diversion model: 109,322 displays on searches for the mark between 2017 and 2021, 7,452 clicks, a 6.82% click-through rate, and 236 callers who mentioned the competitor — 0.216% of total exposures. "No reasonable jury would conclude that this percentage is anything but de minimis," the court held, adding that the bolded "Ad" designation "sufficiently distinguishes ALG's advertisements from the search's organic results." Certiorari was denied on 27 May 2025.

What the Google Ads record actually contains

The auction insights report compares an advertiser's performance against others competing in the same auctions. Google's own definitions: impression share is "the number of impressions you received divided by the estimated number of impressions you were eligible to receive"; overlap rate is "how often another advertiser's ad received an impression when your ad also received an impression"; outranking share is "how often your ad ranked higher in the auction than another advertiser's ad, or if your ad showed when theirs didn't"; and the top of page and absolute top of page rates measure placement above the unpaid results.

The search terms report shows the actual queries that triggered an ad, with the matched keyword, the match type — exact, phrase, broad, or a close variant — and performance metrics.

Impression share generally is impressions divided by total eligible impressions. Google warns that processing takes twenty-four to forty-eight hours, that a dash is shown where data is insufficient, and that "small fluctuations over time don't necessarily indicate that action is needed," because bid, quality and system changes alter the set of auctions an advertiser is estimated to have been competitive in.

What each metric proves about diversion, and what it does not

These limits belong in the report, stated out loud, before an opponent states them for you.

• Auction insights discloses no competitor bid amounts, no competitor keywords, and no absolute impression volumes. It is share-based and relative. It shows that two advertisers were repeatedly in the same auctions and who tended to place higher, not how many people saw either ad. Google does not make it available where impression share is under 10%.
• The search terms report is privacy-thresholded. Google describes it as a list of terms "a significant number of people have used." Low-volume queries are excluded by design. The absence of a competitor's brand term is therefore not evidence that the term never triggered the ad — a point that cuts against whichever party relies on absence.
• Impression share estimates a denominator nobody can observe. "Eligible impressions" is modeled, not counted.
• A rejected trademark complaint to Google proves nothing about the law. Google does not investigate complaints about keywords at all; it restricts trademark use in ad text and display URLs by competitors, and does not restrict landing pages primarily dedicated to selling the corresponding goods or providing informative details about them.

The rest worth requesting: clicks and click-through rate on the accused impressions; cost per click and total spend on the disputed keywords, a fair proxy for the value the defendant placed on that traffic; conversions and conversion value, with the attribution model and lookback window identified, since last-click and data-driven models produce materially different figures; the plaintiff's own click-through rates on the same queries as a baseline; and call-tracking records where a distinct number was used.

The step that gets a diversion model excluded

Everything above establishes exposure. None of it establishes substitution — that a click which went to the defendant would otherwise have gone to the plaintiff. That assumption is the most attacked step in a keyword damages model, because it is usually assumed rather than supported.

The 2024 Ninth Circuit record is the cautionary illustration. Against 109,322 exposures and 7,452 clicks, the identifiable confused callers numbered 236. A model treating those clicks as diverted customers would have overstated the confused population by a factor of more than thirty, and the court called the actual rate de minimis as a matter of law.

What a technical expert can defensibly supply is the exposure and behavior layer: which queries triggered which advertisements, how many times, with what click-through, at what cost, landing where, and how the plaintiff's own results performed over the same period. What converts that into a damages figure — a substitution rate, a margin, a counterfactual — is an economist's work, and the division should be explicit in both reports. In my experience the keyword opinion that fails is not the one that got the search facts wrong. It is the one that quietly carried a substitution assumption inside a technical exhibit, where no economist had ever signed off on it.

Frequently Asked Questions

Published 2026-08-11