Competitor and Trademark Bidding

Where the reported decisions draw the line

Twenty years of paid-search litigation have converged on one distinction, and it is about the record rather than about motive. As reported in the federal appellate decisions, buying a competitor's trademark as a keyword is not, standing alone, infringement in the United States. Putting the mark into the visible ad — headline, description, display URL — is where the reported claims start to survive.

The most recent appellate voice is 1-800 Contacts, Inc. v. JAND, Inc., No. 22-1634 (2d Cir. Oct. 8, 2024). The court held that the mere purchase of a competitor's trademarks in the context of keyword search advertising does not constitute infringement, and that the correct comparison for the similarity-of-the-marks factor is what appears in the advertisement, not the keyword purchased. With the mark absent from the ad, dissimilarity dominated even though the trademark owner had the better of several other factors.

Whether particular conduct is actionable is a question for counsel; what follows is what the account record shows. That is most of what those decisions treat as decisive: what the ad said, what the assets said, what was targeted, and on which dates — each a preserved artifact with a retention clock running on it.

How a mark gets into an ad nobody drafted

The highest-yield first check is the sentence that turns up in almost every response to a demand letter:

We never put their trademark in an ad.

On the record that is frequently untrue, and untrue without anyone having lied. Several mechanisms put a competitor's mark into served creative with no human drafting it.

• Dynamic keyword insertion. DKI substitutes the matched query into the headline. If a broad match keyword matched a competitor-brand query and DKI was enabled, the served headline carried the competitor's mark. That is visible from the ad-level record joined to the search terms report.
• Ad customizers and automatically created assets. Text assembled by the platform or from a feed rather than written by whoever approved the campaign.
• Sitelinks, callouts and structured snippets. Assets are ad text for policy purposes, and the part of the creative most often missed by a request that asks only for "the ads."

So the request worth drafting is not "produce your ads." It is every ad and asset at the asset level with the dates each was eligible, the search terms report for the same period, and the change history for the ad groups involved.

What the case law holds, and what it does not

The table is the operative rule as the reported decisions state it — holdings summarized, not advice on any set of facts.

Three cases are cited constantly for things they did not decide. Rescuecom Corp. v. Google Inc., 562 F.3d 123 (2d Cir. 2009) held only that selling a mark as a keyword trigger clears the Lanham Act's "use in commerce" threshold, and settled in 2010 with no merits ruling; Rosetta Stone Ltd. v. Google, Inc., 676 F.3d 144 (4th Cir. 2012) vacated a summary judgment for Google on fact questions about intent and actual confusion, and also settled with no liability finding. Neither is a merits holding. Multi Time Machine, Inc. v. Amazon.com, Inc., 804 F.3d 930 (9th Cir. 2015) is an on-site search results case, not a sponsored-search case; its value here is that clear labeling can defeat confusion as a matter of law.

The divide shows most cleanly in 1-800 Contacts, Inc. v. Lens.com, Inc., 722 F.3d 1229 (10th Cir. 2013): the keyword-bidding claim failed, while the claim over affiliate ads carrying the mark in the ad text survived on contributory liability. Same defendant, same opinion, opposite outcomes, and the variable was whether the consumer saw the mark. The opinion in JAND is the shortest route to the current framing.

Initial interest confusion, circuit by circuit

Initial interest confusion — the theory that using a mark to capture attention infringes even if the confusion clears before purchase — is often called dead. Narrowed unevenly is more accurate:

Whatever weight it still carries now depends heavily on whether the ad told the consumer who was advertising. That is a question about served creative and screen position, and it is discoverable rather than argued.

Google's trademark policy decides nothing about infringement

Google's trademark policy is a private set of rules about what may run on Google's own inventory. Both sides tend to read it as though it were law:

• Keywords are not restricted. Google states it will not restrict or investigate the use of trademarks as keywords, so a complaint is not a route to blocking keyword targeting.
• Ad text is restricted where the ad comes from a direct competitor, or the mark is used in a confusing, deceptive or misleading way — subject to exceptions for resellers, informational sites and descriptive use.
• Complaints run through Google's legal troubleshooter, are accepted only for countries and industries where the owner has demonstrated rights, and restrictions attach to ads using the same second-level domain in the final URL.

One date changes what a restriction means. On July 24, 2023 Google moved worldwide from industry-wide restriction to advertiser-specific and ad-specific restriction, phasing out pre-existing industry-wide restrictions over a stated 12 to 18 months. A restriction in force before mid-2023 and one after are not the same fact. Google's Trademarks policy page was read on August 14, 2026.

A restriction imposed or refused shows what Google's policy team decided, not what a court would. The systems are independent both ways: policy compliance is not a defense to infringement, and a policy violation is not proof of it.

The FTC matter was antitrust, not trademark

The item mis-stated more often than any other here is the 1-800 Contacts FTC proceeding, usually in a way that would hurt the party relying on it. Between roughly 2004 and 2013, 1-800 Contacts entered agreements with about fourteen rival contact-lens sellers, most settling trademark suits over search advertising: both sides agreed not to bid on the other's marks and to employ negative keywords so their ads would not appear on searches for the other's brand terms. The FTC alleged this restrained truthful advertising and price competition under Section 5 of the FTC Act, and the Commission found for complaint counsel.

In 1-800 Contacts, Inc. v. Federal Trade Commission, 1 F.4th 102 (2d Cir. June 11, 2021), the court granted the petition for review, vacated the order and remanded with instructions to dismiss. The Commission had applied the wrong standard by treating the agreements as inherently suspect, when trademark settlements could plausibly be thought to have a net procompetitive effect; the rule of reason applies; and proof of anticompetitive effect was inadequate, notably for the absence of an empirical analysis of the effect on contact-lens prices. The slip opinion is on govinfo.

What it does not do is bless mutual no-bid clauses categorically. It is an antitrust decision reviewing an agency order, it disturbs no trademark law, and it binds one circuit.

What I can establish from the account record

The examination is narrow, mechanical and reproducible. From a full account export and the ordinary reporting suite, I can state as fact:

1. What was targeted. The keyword list with match types and the dates each was added, paused or removed, and the negative keyword lists that did or did not exclude the competitor's brand terms.
2. What served. Every ad and asset, its text, its ad group and the period it was eligible — including assets added and deleted mid-period, which change history preserves after the asset is gone.
3. What the queries were. The search terms that triggered the ads, joined to cost and clicks, so spend on the competitor's brand queries is separable.
4. Volume and position. Impressions and clicks by ad and by date, and where the ad sat relative to organic results — which Network Automation puts directly in issue.
5. Who changed what, and when. Change history carries user identity and timestamp for keyword, ad, asset and setting changes, including those made through the API, Google Ads Editor and automated rules.

That is a factual record about served advertising. Whether it supports a likelihood of confusion, and what it is worth if it does, are questions the record informs and does not answer.

What the account cannot show

An honest boundary is more useful than an expansive one. Each of these is a place an opposing expert waits.

• Likelihood of confusion. No export establishes it. Clickthrough rate is circumstantial at best, and Lens.com used a low rate as evidence against confusion, which does not make a high rate proof of it. Confusion is ordinarily a survey question.
• Intent to trade on the mark. A keyword list shows targeting, not motive. Intent lives in internal messages, campaign naming and the drafting history of the copy.
• Whether a consumer was diverted. The platform records a click, not a belief. Call recordings, chat transcripts and complaint records are where that sits.
• The exact ad one person saw on one day. Responsive search ads assemble headlines and descriptions per auction. The record preserves the asset set and asset-level serving, supporting a reconstruction of likely combinations; absent a screenshot, "this is the ad the plaintiff saw" is usually not available.
• Lost sales. Impression share lost to a competitor is not a sales figure. Attributing revenue to a rival's bidding requires a counterfactual and is contested.
• What the competitor bid or spent. Auction Insights reports share-of-voice metrics only, never a competitor's bid, budget, cost or clicks, and shows nothing below a 10% impression share.

How long the served-ad record survives

Evidence here is unusually perishable: the creative in dispute may have been deleted the week the demand letter arrived. Deleted ads and assets stay visible in reporting and change history until the window closes; after that they are not hidden, they are gone.

Sources: Google Ads Help, "About change history" and the Google Ads Data Retention Policy, and the API documentation for change_event and change_status, read August 14, 2026; the reporting figures took effect June 1, 2026.

The consequence is a sequencing one. A clean, machine-readable, hashable pull of field-level change data has to happen inside thirty days; after that the interface is the only source, and it is a two-year window that moves daily. If the campaigns ran longer ago, plan on ad text being reconstructed from reporting exports, screenshots and the other side's own files — and expect the completeness of that reconstruction to be challenged.

Frequently Asked Questions

Published 2026-08-14