Trademark Infringement in Organic Search

Where a mark can appear on a page, and how each location is proved

An organic-search trademark claim is almost never about one appearance of the mark. It can sit in a dozen separately provable places in one document, each with a different capture method, a different visibility, and a different evidentiary weight.

• The <title> element — the text usually rendered as the clickable blue link. Google may rewrite it, so the element and what appeared on the results page are two facts needing two exhibits.
• The meta description — a <meta> tag holding a short summary of the page. Not a ranking input, but Google often uses it verbatim as the snippet beneath the link.
• Headings and body copy, captured from the rendered DOM: the page as a browser assembles it after JavaScript runs, which is not always what the server sent.
• The alt attribute on an image, invisible on screen but read by screen readers and used by Google Images.
• Structured data — machine-readable markup describing the page to a parser rather than to a reader.
• Hidden text: same-color-on-same-color, off-screen positioning, or a container collapsed to zero height.

The metatag cases, and how much of the mechanism survives

Brookfield Communications, Inc. v. West Coast Entertainment Corp., 174 F.3d 1036 (9th Cir. 1999) is where this line starts. The Ninth Circuit enjoined use of "MovieBuff" in metatags and in a domain name, holding initial interest confusion actionable even where confusion is dispelled before purchase. Its billboard illustration has been quoted ever since: a competitor's sign reading "West Coast Video: 2 miles ahead at Exit 7" where West Coast is at Exit 8 and the competitor is at Exit 7.

In Playboy Enterprises, Inc. v. Welles, 279 F.3d 796 (9th Cir. 2002), metatag use by the woman who had actually held the title was nominative fair use.

The problem with citing this line today is that all of it concerns <meta name="keywords">, which Google abandoned as a ranking input long ago. The doctrine survives; the mechanism largely does not. A modern equivalent involves hidden <div> text, off-screen CSS, alt attributes, or markup no reader sees — each a question about what a document contained on a date, not a question about rankings.

Initial interest confusion, circuit by circuit

This doctrine is in retreat, and a page written for litigators has to be precise about where.

• Ninth Circuit. Recognized but narrowed hard. Network Automation, Inc. v. Advanced Systems Concepts, Inc., 638 F.3d 1137 (9th Cir. 2011): "the owner of the mark must demonstrate likely confusion, not mere diversion." Reaffirmed on the facts in Lerner & Rowe PC v. Brown Engstrand & Shely LLC, 119 F.4th 711 (9th Cir. 2024).
• Seventh and Tenth Circuits. Applied in Promatek Industries, Ltd. v. Equitrac Corp., 300 F.3d 808 (7th Cir. 2002), adopted in Australian Gold, Inc. v. Hatfield, 436 F.3d 1228 (10th Cir. 2006); the Tenth found the diversion de minimis in 1-800 Contacts, Inc. v. Lens.com, Inc., 722 F.3d 1229 (10th Cir. 2013).
• Eighth Circuit. The most recent to embrace it. Select Comfort Corp. v. Baxter, 996 F.3d 925 (8th Cir. 2021) reversed a jury instruction requiring confusion at the time of purchase, reasoning that the 1962 amendment removing the word "purchasers" permits presale claims, and holding it error to treat consumer sophistication as settled as a matter of law.
• Second Circuit. Presale confusion is recognized generally, but in the keyword context the court directed that mark similarity "should be assessed as it relates to the paid advertisement's appearance on the result page" — 1-800 Contacts, Inc. v. JAND, Inc., No. 22-1634 (2d Cir. 8 October 2024).

Where the doctrine survives, the plaintiff must still prove confusion rather than diversion, which makes the technical assignment establishing what the user saw at the moment of the alleged diversion.

Nominative fair use, and the split that decides which test applies

Where a defendant used the mark to refer to the plaintiff's own goods — a reseller, a repair shop, a comparison page — nominative fair use is usually the first defense raised, and which test applies depends on the circuit.

The Ninth Circuit's model civil jury instruction sets out three elements: that "the product in question was not readily identifiable without use of the trademark"; that "the defendant used only so much of the trademark as was reasonably necessary to identify the product in question"; and that "the defendant did not do anything in connection with the trademark that would suggest sponsorship or endorsement by the plaintiff." The doctrine originates in New Kids on the Block v. News America Publishing, Inc., 971 F.2d 302 (9th Cir. 1992).

Toyota Motor Sales, U.S.A., Inc. v. Tabari, 610 F.3d 1171 (9th Cir. 2010) matters most for domain-level use. Reversing an injunction against buy-a-lexus.com, the court held it enough for necessity "that the Tabaris needed to communicate that they specialize in Lexus vehicles,"

The frameworks diverge from there. In the Ninth Circuit the three-element test replaces the likelihood-of-confusion analysis. The Third Circuit, in Century 21 Real Estate Corp. v. LendingTree, Inc., 425 F.3d 211 (3d Cir. 2005), treats it as an affirmative defense reachable even where confusion has been shown. The Second Circuit has adopted a third framework of its own. The Supreme Court denied certiorari on 9 January 2017 and the split has stood since.

Reconstructing what the page contained on a given date

A crawl of the defendant's site today establishes what is there today and nothing about the period pleaded. The sources that reach backward are limited, and each carries a foundation problem.

• Internet Archive captures. Dated snapshots of raw HTML, including elements no user saw. Admission is contested — at least one circuit has held a court may not take judicial notice of an archived page, because a private archive is not a source whose accuracy cannot reasonably be questioned, which leaves authentication under Rule 901 as the route.
• The party's own systems. CMS revision history, deployment logs, templates in version control, CDN configuration. The best evidence usually lives here, reachable only in discovery.
• Server access logs. They establish that a URL existed, was requested, and returned a given status code — not what the page said.
• Contemporaneous crawls. Where a party ran a crawler such as Screaming Frog during the window and kept the output, the saved file is a dated capture of titles, descriptions and directives sitewide.
• Search Console. The owner's record of which URLs Google indexed and in what state.

Two distinctions decide whether the reconstruction holds. Raw HTML is not the rendered DOM; a mark injected by script appears in one and not the other, and the exhibit must say which was captured. And what the page contained is not what Google displayed, because Google rewrites titles and snippets — so if the claim is that the mark appeared in the results, the results-page capture is the exhibit and the source is corroboration.

Hidden text, and what a technical choice actually proves

Google's spam policies define hidden text and links as "placing content on a page in a way solely to manipulate search engines and not to be easily viewable by human visitors,"

A manual action — a penalty applied by a human reviewer at Google, as distinct from an algorithmic ranking change — is recorded in the site owner's Search Console account under a report named "Hidden text and/or keyword stuffing." If the account shows one for the period in suit, that is a contemporaneous third-party record that the conduct occurred and was classified as manipulation — and it does not depend on the expert's own reconstruction.

What can be said about a technical choice is bounded, and staying inside that boundary is what keeps the opinion useful. A declaration of color:#FFFFFF on a white background, or text-indent:-9999px on a block containing a competitor's mark, has one plausible technical effect: the text is present for a parser and absent for a person. Describe the declaration, describe the effect, stop. The inference that the effect was the purpose belongs to the finder of fact, and an expert who reaches for it hands opposing counsel the cross-examination.

Structured data as trademark use — an unanswered question

Structured data is markup embedded in a page so a machine can read its contents: that this is a product, that this is its brand, that this is its price. It drives rich results, and because it is written for a parser, it can assert something the visible page never says, including a brand name.

Google's structured data policies state: "Don't mark up content that is not visible to readers of the page." That is a platform rule with a platform consequence — a structured data manual action costs rich-result eligibility. It is not a legal standard and should never be offered to a court as one.

I searched for a reported United States decision treating schema.org or JSON-LD markup as trademark use in commerce, or as the basis of a likelihood-of-confusion finding, and did not find one. The metatag line above is the closest analogue and predates structured data by more than a decade. That absence is worth stating plainly. A claim built on brand markup no consumer sees would be a question of first impression in most courts, and the argument that a machine-readable assertion is a use seen by no purchaser cuts both ways. Raise it as an open question, and if an opponent asserts courts have decided it, ask for the citation.

The line between the technical opinion and the legal conclusion

The opinion that survives cross-examination has a visible edge to it. What a search expert can establish, with a source and a date attached to each:

• what content was present at what URL on what date, and in which HTML element;
• whether it was visible to a sighted user or only to a parser, and whether the markup matched the visible content;
• whether the version Google rendered differed from the raw source, and whether the mark appeared in the snippet Google generated;
• how the disputed pages were reachable, and the queries for which they were indexed.

What the same expert cannot establish: whether any consumer was confused, which is survey territory; whether a use was nominative, which is a legal characterization; and whether the intent was to deceive. Intent may be evidenced by a technical choice, and describing the choice and its only plausible effect is legitimate. Characterizing the actor's state of mind is not, and in my experience it is the most common way a competent technical opinion gets discounted.

Frequently Asked Questions

Published 2026-08-11