Does Schema Markup Help You Get Cited by AI?
No — controlled testing shows schema markup produces no measurable lift in AI citations. Ahrefs' 1,885-page difference-in-differences study found no uplift, and a slight decline on AI Overviews. Schema is still valid infrastructure for other reasons.
No — schema markup produces no measurable lift in AI citations. The most rigorous public test to date found that adding structured data did not raise citation rates on any AI platform, and slightly lowered them on Google AI Overviews. Schema remains useful infrastructure — but not as a citation lever.
Verdict
Adding schema markup does not increase how often AI engines cite you. In Ahrefs' controlled 1,885-page test, the effect was statistical noise on AI Mode and ChatGPT and a small negative on AI Overviews. Keep schema for rich results and entity clarity — not for AEO citations.
What does the controlled evidence show?
The controlled evidence shows no positive effect of schema on AI citations. In May 2026, Ahrefs published a study that tracked 1,885 pages which added JSON-LD schema between August 2025 and March 2026, matched against roughly 4,000 control pages, and ran a difference-in-differences analysis (plus three other statistical tests) on citations 30 days before versus after the markup was added. This is the gold standard for isolating cause: it compares pages that changed against similar pages that didn't.
| Platform | Change after adding schema | Interpretation |
|---|---|---|
| Google AI Mode | +2.4% | Statistical noise |
| ChatGPT | +2.2% | Statistical noise |
| Google AI Overviews | −4.6% | Significant — slight decline |
The small positives on AI Mode and ChatGPT sit close enough to zero that Ahrefs classifies them as noise across thousands of URLs. The AI Overviews decline was statistically significant (odds of chance roughly 1 in 2,500). The conclusion was blunt: "adding schema produced no major uplift in citations on any platform."
Why doesn't schema move AI citations?
Schema doesn't move AI citations because engines extract from the visible, rendered text — not the structured-data block in your page's head. Analyses of AI crawler behavior, including work by searchVIU, find that AI crawlers parse the human-readable HTML a reader actually sees, then retrieve and quote passages from it. Your JSON-LD describes the page to traditional search features, but the sentence an engine lifts comes from your prose.
The mechanism, not the markup
An answer engine cites the passage that best answers the question. That passage lives in your visible content — which is why extractability (answer-first, self-contained writing) is the lever, and markup is not. Schema can tell a parser "this is an FAQ"; it can't make a buried answer quotable.
This also fits the broader evidence: the Princeton GEO study found the tactics that lift AI visibility are citations, quotations, and statistics — content qualities, not metadata. Markup wasn't the differentiator.
What is schema markup actually good for?
Schema markup is genuinely valuable — just not for AI citations. It remains a core part of technical SEO and entity clarity, and you should keep using it for the things it actually does.
- 1
Rich results in traditional search
Schema powers rich snippets, review stars, FAQ accordions, breadcrumbs, and other enhanced listings in classic search results.
- 2
Entity understanding & the Knowledge Graph
Structured data helps engines connect your pages to defined entities (organizations, people, products) — supporting how you're recognized.
- 3
Unambiguous machine parsing
Schema removes ambiguity about what a page is — useful wherever a system needs structured facts rather than prose.
- 4
Feature eligibility
Some search features require valid markup to appear at all. Schema is the price of admission for them.
This is exactly the position The AEO Canon takes and our guide to structured data for AEO details: schema is valid infrastructure, not a proven citation lever. Implement it correctly, then stop expecting it to raise your AI visibility.
What should you do instead?
Instead of relying on markup, invest in the levers controlled testing actually supports: answer-first, self-contained passages (extractability); inline statistics, quotations, and named sources (credibility); genuine off-site authority; and freshness. Keep your schema clean for the rich results and entity clarity it earns — and put your AEO effort where the evidence points.
This is one of several widely-believed tactics that don't survive controlled testing — see the full set in AI search optimization myths, debunked, and the foundations in what is AEO.
Frequently asked questions
- Does schema markup help you get cited by AI?
- No — there's no measurable citation lift. Ahrefs' controlled study of 1,885 pages that added JSON-LD schema found no uplift in AI citations on any platform, and a small but statistically significant decline on Google AI Overviews. Schema is valid infrastructure for rich results and entity clarity, but it is not a proven AI-citation lever.
- Why doesn't schema help AI citations?
- Because answer engines primarily read the visible, rendered text of a page, not the structured-data blocks in the head. searchVIU and others have found AI crawlers parse the human-readable HTML — the same passages a reader sees — so the answer-first prose, not the JSON-LD, is what gets extracted and cited.
- Should I remove schema markup then?
- No. Schema still earns rich results in traditional search, helps engines and the Knowledge Graph understand your entities, and powers features like FAQ and HowTo eligibility. Keep it as low-cost infrastructure — just don't expect it to raise your AI citation rate.
- What actually helps with AI citations instead?
- Answer-first, self-contained passages; inline evidence (statistics, quotations, named sources); genuine off-site authority; and freshness. Those are the levers controlled testing supports — not markup.
Last updated .