- This site's dedicated schema-markup piece already covers the pooled result: no measurable AI-citation gain across schema types overall, and a small, statistically significant decline on Google AI Overviews. This post is the type-by-type breakdown that pooled result leaves open.
- FAQPage schema's AI-citation lift is contested: a vendor figure claiming a 2.7x improvement (from a 50-domain test) doesn't hold up against an independent SE Ranking study finding a much smaller, roughly 11% difference.
- Product and Review schema populated with real pricing, ratings, or specifications is the one type with a statistically significant, isolated citation-rate effect, not because of the tag, but because it guarantees the exact data a model is looking for is present and structured.
- Organization schema with sameAs links to authoritative profiles (Wikidata, Wikipedia, LinkedIn) does real entity-resolution work regardless of its AI-citation effect, which is a separate and independently justifiable reason to have it.
- The practical sorting rule: if a schema type exists to carry data a model would otherwise have to infer (price, rating, spec, FAQ answer), it's worth using. If it exists mainly to label a page's genre (Article, generic Organization with no sameAs), it's doing less work than it looks like it's doing.
Does schema markup help AI citations? already walks through the core, pooled finding: a matched, controlled study found no measurable AI-citation benefit across schema types as a group, and a separate study isolated a real exception once it stopped pooling. This piece picks up exactly where that leaves off: a type-by-type look at which specific schema actually shows evidence behind it, and which is mostly carrying its own weight on faith.
Why does pooling schema types together hide the real picture?
Ahrefs' own 1,885-page matched study, the most rigorous controlled test available on this question, pooled every schema type together, Article, FAQPage, Product, HowTo, and Organization, rather than isolating which type was added. The study's authors flagged this directly as an open question their own data couldn't resolve: "it's possible some types help more than others." Averaging a type with a real effect together with four types that show none produces exactly the "no net effect" result the pooled study found, without telling you which type is actually worth the effort.
What does the evidence say about FAQPage schema specifically?
This is where the gap between vendor marketing and independent verification is widest. One vendor test, run across 50 domains, claims FAQPage markup produces roughly 2.7 times more AI citations. An independent SE Ranking study, with a larger and more disclosed methodology, found a much smaller difference: pages with FAQ schema averaged around 4.9 citations against 4.4 for pages without it, a real but modest gap of roughly 11%, not 170%.
Those two figures aren't measuring noise in the same direction; they're an order of magnitude apart, which is a strong signal that the 2.7x number is the kind of unverified vendor stat already flagged elsewhere on this site as something to trace back before repeating. Treat FAQPage schema as plausibly worth a small amount of credit, not as the dramatic lift it's frequently marketed as. One more wrinkle worth knowing: Google retired the visible FAQ rich result from its own search results pages in 2026, though the schema type itself remains a valid part of the Schema.org vocabulary and isn't the same thing as whether it influences AI citation specifically.
What does the evidence say about generic Article and Organization schema?
Weak to none, independent of the pooled study's overall null result. The Fischman working paper that isolated schema by type found generic schema types like Article and plain Organization markup cited at 41.6%, essentially the baseline rate, against 61.7% for attribute-rich Product/Review schema. The mechanism, confirmed separately by a peer-reviewed SIGIR 2026 study, is that formatting-only edits (which is what Article and bare Organization schema fundamentally are: labels describing a page's genre) carry little independent weight. A model isn't looking for confirmation that a page is "an Article." It's looking for a fact to extract.
What does the evidence say about Product and Review schema?
This is the one type with a real, statistically significant, isolated effect: 61.7% citation rate for pages using Product/Review schema populated with actual pricing, ratings, or specifications, against 41.6% for generic types, a gap that held up at p = .012 across 730 analyzed AI citations. The mechanism is specific and doesn't generalize to schema broadly: Product/Review schema is structurally built to carry exactly the data points (price, aggregateRating, spec fields) that a separate, peer-reviewed study independently identified as citation gatekeepers. The schema wrapper earns its keep here because there's a fact inside it worth extracting, not because the wrapper itself is special.
What about Organization schema's sameAs links specifically?
This is worth separating from the "does it move citations" question entirely, because it's doing a different job. Organization schema with sameAs properties pointing to a brand's Wikidata item, Wikipedia page, and other authoritative profiles is a standard entity-resolution signal: it's one of the ways a brand tells search engines and knowledge graphs "these different records all describe the identical entity." That's the same disambiguation problem covered in what is entity SEO and what is entity consistency, and it's worth doing for that reason independent of whether it ever moves an individual AI citation. Conflating "does this help entity resolution" with "does this help AI citation" is exactly how a type like Organization schema gets oversold on the wrong metric.
So which structured data types are actually worth the effort?
A practical sorting rule, based on everything above: schema that's structurally built to carry a fact a model would otherwise have to infer or skip (Product/Review with real pricing and ratings, FAQPage with a direct, modest and contested lift) earns its implementation effort for AI citation specifically. Schema that mainly labels a page's genre (plain Article, bare Organization with no sameAs linking) does close to nothing for AI citation on the evidence available, though Organization schema with sameAs links still earns its keep for entity resolution, a separate and real job.
| Schema type | Measured AI-citation effect | Why |
|---|---|---|
| Product/Review (populated) | Real, statistically significant (61.7% vs. 41.6%) | Carries exact data (price, rating, specs) a model is looking for |
| FAQPage | Small and contested (roughly 11%, not the marketed 2.7x) | Direct Q&A format helps modestly; vendor figures for it are inflated |
| Article (generic) | None measured | Formatting-only; no gatekeeper data attached |
| Organization (bare) | None measured for citation; real for entity resolution | Labels identity, doesn't carry citable facts on its own |
| Organization + sameAs | Not measured for citation; real for entity resolution | Links disambiguate the entity across sources, a separate job |
None of this contradicts the pooled Ahrefs finding; it explains it. A null result across every type, averaged together, is exactly what you'd expect when one type (Product/Review, populated) carries a real effect and the rest carry none.
Frequently asked questions
Does adding structured data help you get cited by AI models?
Pooled across all types, the best controlled study available (Ahrefs, 1,885 matched pages) found no measurable gain, and a small, statistically significant decline on Google AI Overviews specifically. Broken down by type, the picture is less uniform: generic types show nothing, and Product/Review schema carrying real data shows a real, separate effect.
Does FAQPage schema help AI citations?
The evidence is thin and contradictory. One vendor test of 50 domains claims a 2.7x lift; an independent SE Ranking study found a much smaller difference, about 4.9 citations against 4.4 for pages without an FAQ block, roughly 11%, not 170%. Treat the larger figure as unverified marketing and the smaller one as the more credible, though still modest, signal.
Which structured data type has the strongest evidence behind it?
Product and Review schema populated with real, concrete fields (price, aggregateRating, specifications), per a working paper that isolated schema type and found a statistically significant citation-rate gap versus generic schema types. This is already the main finding of this site's dedicated schema-markup piece; this post adds the type-by-type breakdown that finding implies.
Should every page carry Organization schema regardless of AI citation effects?
Yes, for a different reason than AI citation specifically. Organization schema (and sameAs links to Wikidata, Wikipedia, and other authoritative profiles) is one of the standard entity-resolution signals feeding knowledge panels and a brand's resolvability as a single entity, a job it does independent of whether it moves any individual AI citation.
Sources referenced
- Ahrefs: We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved. (May 2026)
- Fischman, "Does Schema Markup Predict AI Citation? A Cross-Platform Empirical Study of Structured Data and Generative Engine Optimization" (SSRN working paper, Feb 2026)
- Vishwakarma, Kumar & Jamidar, "What Gets Cited: Competitive GEO in AI Answer Engines" (SIGIR '26)
- SE Ranking: FAQ schema and AI citation rate analysis
- Google Knowledge Graph Search API documentation
This piece is the type-level follow-up to does schema markup help AI citations?. For the entity-resolution job Organization schema and sameAs links do independent of citation effects, see what is entity SEO and what is a knowledge graph.
