ClaimReview, and the schema your competitors won't ship.

Snopes uses ClaimReview. Reuters Fact Check uses ClaimReview. PolitiFact, FactCheck.org, AFP Fact Check, Lead Stories, and the Poynter network all use ClaimReview. The schema was published by Schema.org in 2016 and adopted by Google's fact-check program the same year. It is the format that powers the small "Fact Check" widgets that appear under contested news results.

No DTC brand uses ClaimReview. The Schema.org documentation does not even contemplate the product-marketing use case. Search "ClaimReview ecommerce" in May 2026 and the first page of results is unanimous on this point: ClaimReview is for journalism, not for product pages.

This essay argues the opposite. A product page that says "tested by 200 dermatologists in 2025" is making a claim. The claim is either true or it is not. A schema designed to attach a verdict and a date to a falsifiable claim is exactly the schema that copy should be sitting under. The brands that ship it first will be the brands the engines learn to trust.

ClaimReview is a Schema.org type. Its required fields are short.

`claimReviewed`: the text of the claim. A sentence, in quotation marks, lifted from somewhere. `itemReviewed`: a structured reference to the original claim, including who said it and when. `reviewRating`: a rating, on whatever scale the reviewer uses, with a numeric value and a label. `author`: the entity doing the review, with a name and a URL. `url`: where the review lives. `datePublished`: when the review was published.

Snopes uses it like this. The claim ("the Federal Reserve has been abolished") gets a rating ("False"), an author ("Snopes"), a date, and a URL. A reader on a Google results page sees a small box: claim, verdict, source. The schema makes the box.

Read the field list again and notice what it does not require. It does not require that the author be a journalism organisation. It does not require that the rating be on a five-point truthfulness scale. It does not require that the claim be political. It requires only that some entity, somewhere, has reviewed a specific quoted claim and produced a verdict.

A brand can be that entity. A lab can be that author. The verdict can be "Substantiated" instead of "True." The schema does not care.

Read a serious DTC product page. Skin care, supplements, hardware, footwear. The copy is dense with claims that can, in principle, be checked.

"Tested by 200 dermatologists in 2025." That is a claim. Either a study with N=200 happened in 2025 or it did not.

"Reduces fine lines by 27% over eight weeks." Either a clinical or consumer panel produced that number or it did not.

"Made in our family-owned factory in Vicenza, Italy, since 1962." Either the factory exists, has been in continuous operation since 1962, and is family-owned, or it is not.

"Vegan, cruelty-free, certified by Leaping Bunny." Either there is a current Leaping Bunny certification or there is not.

These are falsifiable claims. The brand making them has, almost always, an underlying document: a study PDF, a certificate, a factory inspection, a press article. The document is real. The product page just doesn't say which sentence the document substantiates.

ClaimReview is the schema that says.

{
  "@type": "ClaimReview",
  "claimReviewed":
    "89% reported visible
     improvement, N=184.",
  "reviewRating":
    "Substantiated",
  "author":
    "Independent CRO"
}

Imagine a serum product page that contains, inline in the copy, the sentence: "In an independent eight-week consumer study of 184 women aged 30 to 55, 89% reported visible improvement in skin texture."

The brand has the underlying study. It was conducted in March 2025 by an independent CRO. The methodology PDF is hosted at `https://[brand]/research/2025-spring-study.pdf`.

A ClaimReview block on that product page, in JSON-LD, looks like this.

```json { "@context": "https://schema.org", "@type": "ClaimReview", "datePublished": "2025-04-12", "url": "https://[brand]/products/serum-01#claim-texture-89", "claimReviewed": "In an independent eight-week consumer study of 184 women aged 30 to 55, 89% reported visible improvement in skin texture.", "itemReviewed": { "@type": "Claim", "appearance": { "@type": "WebPage", "url": "https://[brand]/products/serum-01" }, "datePublished": "2025-04-08", "author": { "@type": "Organization", "name": "[Brand]", "url": "https://[brand]" } }, "author": { "@type": "Organization", "name": "[Independent CRO]", "url": "https://[cro-domain]" }, "reviewRating": { "@type": "Rating", "ratingValue": "Substantiated", "alternateName": "Substantiated by independent study", "ratingExplanation": "Eight-week consumer study, March 2025, N=184. Methodology: https://[brand]/research/2025-spring-study.pdf" } } ```

That block does four things at once.

It quotes the exact sentence appearing on the page. The claim is precisely the marketing line, not a paraphrase.

It identifies who said it (the brand) and who reviewed it (the CRO). The two entities are distinct, which the reasonable-consumer test cares about and the engines care about.

It cites a methodology PDF. The PDF is the document the schema would be lying about if the underlying study did not exist. The risk of attaching the schema to a non-existent study is therefore exactly the risk of attaching your name to a false footnote in any other publication: a category of risk most brands have not, historically, taken at the page level.

It assigns a date. The 2025-04-12 date stamps the review. An engine reading the page in 2026 can decide how much to discount the claim for age. A brand that re-validates the study annually can simply update the date.

The schema is technically free. JSON-LD is a script tag. A Shopify theme dev can add it in an afternoon. Yotpo, Okendo, Bazaarvoice, Junip, Stamped, Loox, and Reviews.io all have the technical capacity to ship ClaimReview templates for their customers tomorrow.

None of them will. Three reasons.

First, the schema requires an author entity distinct from the brand. The author has to be the reviewer, not the seller. ClaimReview asserts that a third party has reviewed the claim and produced a verdict. A platform that ships ClaimReview-as-a-feature would have to either become that third party (which it is not, it is a review-distribution vendor) or instruct its customers to identify a real independent reviewer (which most customers do not have).

Second, the schema makes the claim newly liable. A brand currently writing "89% reported visible improvement" in body copy carries a marketing-claims liability under FTC Section 5 and, since October 2024, under the Consumer Reviews and Testimonials Rule. The text is the liability surface. Adding a structured-data block that quotes the exact sentence does not increase the legal liability (the sentence is already on the page) but it increases the visibility of the liability. An adversarial reader can find every falsifiable claim on the brand's site with a single regex against the JSON-LD. Many brands would, on advice of counsel, prefer that those claims remain in prose.

Third, the schema rewards brands with real documents. A brand that has actually run the study, has the certificate, has the factory inspection, benefits enormously from ClaimReview. A brand whose copy is aspirational benefits from its absence. A platform that ships ClaimReview-as-a-default would force a sorting of its customer base into the substantiated and the unsubstantiated. Few platforms will choose to do that sorting on their customers' behalf.

The result is that ClaimReview, as a competitive surface, is wide open. Any brand that ships it ships a schema the engine reads but the competing brand has not even attempted.

GPTBot, ClaudeBot, and PerplexityBot all parse JSON-LD. Their public documentation states the parsing; their behaviour confirms it. Schema-described claims appear in citations at higher rates than unstructured ones, per the Princeton GEO paper (December 2024) and the Ahrefs corpus study (March 2026).

ClaimReview specifically has not, as of May 2026, been documented as a citation-multiplier signal by any AI lab. The schema is too rare in product context for the labs to have measured it. But the structural inference is straightforward. The engines have published, repeatedly, that they prefer evidence over assertion. ClaimReview is a literal evidence schema. A claim wrapped in ClaimReview, attached to a methodology PDF, dated, and attributed to a distinct reviewer, is structurally indistinguishable from a fact-checked claim in journalism. The engine has learned to trust those.

The brand running this experiment first will produce a small dataset that the rest of the category will eventually have to read. The dataset will be: did pages with ClaimReview attached to product claims see citation lift in answer-engine summaries? Plausibly yes. Almost certainly not by less than the lift on `Review` schema, which the same engines have weighted positively for years.

The way to know whether ClaimReview is worth shipping is to count.

Open the brand's top ten product pages. Read the body copy. List every claim that is, in principle, falsifiable. A claim about ingredients ("contains 2% retinol"). A claim about provenance ("made in Vicenza"). A claim about testing ("dermatologist-tested"). A claim about endorsement ("worn by Olympic athletes"). A claim about certification ("Leaping Bunny certified"). A claim about results ("89% reported improvement").

For each claim, write down the underlying source. The COA from the lab. The factory address. The dermatologist panel report. The athlete's signed endorsement contract. The Leaping Bunny certificate number. The consumer study PDF.

A claim with a source is a candidate for ClaimReview. A claim without a source is a candidate for deletion.

Most brands, on this audit, will find three categories of claim on their pages. A third have a clean source attached. A third have a source that exists but has not been linked, dated, or made auditable. A third are aspirational copy that nobody could substantiate if asked. The audit takes a few hours. It pays in both legal and citation directions.

The first third can ship ClaimReview tomorrow. The second third can ship it as soon as the document is published or linked. The third needs to be rewritten or removed regardless of any schema decision.

ClaimReview's structure mirrors, almost field-for-field, the documentation that an FTC enforcement defence would need. In an enforcement action under the Consumer Reviews and Testimonials Rule or the broader endorsement guides, the Commission asks: what was the claim, who said it, when did they say it, what evidence substantiates it, who reviewed the evidence. Those are exactly the five fields ClaimReview requires.

A brand running ClaimReview at scale on its product pages is, as a side effect, building an audit-grade record of its own claim substantiation. The JSON-LD is a public-facing artifact. The methodology PDFs the JSON-LD points to are the substantiating documents. The combination is, in regulatory terms, a defensible posture. The brand can demonstrate, on demand, what it claimed, when it claimed it, and what evidence supported the claim at the time.

This is not the reason to ship the schema. But it is a reason not to fear it. The brands that already have substantiation files in order benefit from making them machine-readable. The brands that do not have substantiation in order have a different problem, and ClaimReview surfaces it. Either outcome is a forcing function the brand can use.

Most brands will not run methodology studies. ClaimReview, in its full form, requires either a study or an external certification. That is fine. It is meant to be a high bar.

The narrower bar is the date and the source. A brand whose product page says "since 1962" can attach a ClaimReview block that quotes the sentence, names the brand as itemReviewed.author, names an external journalist or trade-press article as the ClaimReview.author, and rates the claim "Substantiated by [outlet] [year]." The author entity has to be real. The article has to be linked. The link has to resolve. That is the bar.

A brand whose product page says "as seen in Vogue" can attach a ClaimReview to the sentence with the Vogue article URL in the methodology field. The verdict, "Substantiated by Vogue, [date]," is exactly what the buyer is being told to believe anyway. The schema makes the citation machine-readable.

A brand that has nothing to put under any claim should, on this reading, write fewer claims. That is also a useful outcome.

A schema designed to attach evidence to a claim has been sitting in the Schema.org vocabulary for nine years. Every product page in DTC commerce is making claims. The two facts have not met. The intersection is not a feature to wait for from Yotpo or Bazaarvoice. It is a fifty-line script tag and a folder of methodology PDFs.

The brand that ships this first does two things at once. It tells the answer engine that its claims are substantiated, in the exact vocabulary the engine has been trained to read. And it tells the regulator that its claims are footnoted, in the format the reasonable consumer test would accept on the witness stand. Both readers, again, are reading the same page.