Does Structured Data Actually Help You Get Cited?

Not as a general rule - and that's now a somewhat different claim than it was a few months ago, because direct research finally exists to check it against. Google says structured data isn't required for its generative Search features and that no special AI-specific schema is needed. That's not the same as Google saying schema has no effect at all, and it's worth being precise about the difference. A real 2026 study now backs up the practical conclusion anyway: adding generic schema to a page that's already being cited doesn't produce a broad citation lift, and for Google's AI Overviews specifically, it was measurably associated with a small decline. Page speed gets a similarly imprecise treatment across most AEO content, and it deserves the same separation this page gives schema: ordinary load speed and severe fetch failure are not the same problem, and only one of them has real evidence behind it.

What Structured Data Actually Does

Structured data gives Google explicit, machine-readable information about a page, its entities, and their attributes. Google can use that information to understand the page and, for supported types that meet its guidelines, make it eligible for richer Search appearances - product price and availability, recipe details, or event information, for instance. Understanding and eligibility are related but separate: providing the markup doesn't guarantee the rich result, since eligibility also depends on the specific type, required properties, Google's own policies, and its display decisions. Google's Article structured-data documentation defines properties for headline, publication and modification dates, and author identity - and it specifically says an author URL or a sameAs link can help disambiguate an author, while the date properties provide more accurate publication information to Search. None of that is nothing. It's also not the same claim as "structured data gets you cited by an AI system" - a different outcome, evaluated by a different process.

What Google Has Actually Said - and Hasn't Said

Worth quoting exactly rather than paraphrasing: Google's own AI-optimization guidance states "structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." That's clear evidence that schema isn't a prerequisite and that there's no special AI-only schema requirement. It is not a statement that ordinary structured data can never contribute to anything, directly or indirectly - and treating it as a blanket denial overstates what Google actually said. The more precise claim, and the one this page actually defends, is narrower: generic schema is not required for Google's generative-search features, and - as of a real 2026 study, not just an absence of evidence - the best available research does not show a general citation uplift from adding it either.

What the Actual 2026 Research Found

Ahrefs ran the most directly relevant study available: a matched difference-in-differences analysis of 1,885 pages that added JSON-LD schema, compared against roughly 4,000 control pages. One detail matters for how far this generalizes: every page in the dataset already had 100+ AI Overview citations before schema was added, so this measures whether schema helps an already-visible page get cited more - not whether schema helps an uncited page get discovered in the first place. Schema types were pooled together (Article, FAQ, Product, HowTo, Organization) rather than tested individually. The results, measured 30 days before and after: Google AI Overview citations fell 4.6% - a small but statistically significant decline, not a null result - while AI Mode (+2.4%) and ChatGPT (+2.2%) both moved by amounts statistically indistinguishable from zero. Read plainly: adding generic schema to a page that's already being cited doesn't produce a citation lift on any of the three platforms tested, and on Google's AI Overviews specifically, it's associated with a small negative effect.

A separate, much smaller preprint adds a genuine nuance, though it deserves real caveats before it changes anything you do. Kurt Fischman's analysis of 730 ChatGPT and Gemini citations across 75 commercial queries and 1,006 pages found no significant effect from schema presence, entity richness, or schema-query alignment generally - but did report that pages with Product or Review schema carrying concrete populated attributes (price, rating, specifications) were cited at higher rates. Two things worth knowing before weighting that finding: it's a non-peer-reviewed preprint with a comparatively small sample, and its author is the founder of a company in the AI-agent-operations space - a direct commercial stake in this exact question. That doesn't make the finding wrong, but it does mean this is a hypothesis worth watching, not a settled result on the level of the larger Ahrefs study. If you sell products or run reviews, populating accurate Product/Review schema is worth doing anyway for its established Search benefits - treat any specific AI-citation lift as unconfirmed.

Does Page Load Speed Affect Whether You Get Cited?

This question actually collapses four different things that deserve separate answers: Core Web Vitals scores, ordinary page-load time, client-side rendering, and outright fetch or timeout failure. Google's own documentation confirms Core Web Vitals are used by its ranking systems, but explicitly states that good scores don't guarantee top rankings on their own - and there's no established Core Web Vitals threshold or ordinary load-speed benchmark shown to independently increase AI citations. Fetch reliability is a genuinely different, more defensible concern. Industry data reported via iPullRank, credited to Profound's analysis of roughly 700,000 pages, found that pages with fetch-failure rates above 75% received roughly 18 times fewer OpenAI citation interactions - in many cases none at all. That data isn't independently published or peer-reviewed, so treat it as preliminary, directional evidence rather than a settled figure. But the underlying logic is sound and matches how retrieval works: if a bot is blocked, repeatedly times out, or can't access the substantive content, the page may never enter the candidate set at all. That's an access and retrieval problem - not proof that making an already-accessible page marginally faster creates a citation preference.

The JS-crawler-access pillar - rendering and fetch reliability, covered in full →
Confirm your own site is actually reachable by AI crawlers →

So Should You Bother With Schema At All?

Yes - for the reasons that were always true, and one narrow new one. Valid Article schema with real author identity, a genuine publish date, and clean entity markup helps Google understand and display your content correctly. That's a real reason to use it, separate from any AI-citation claim. Product and Review schema with real, populated attributes is worth using for its established Search benefits regardless, and the Fischman preprint's finding - heavily caveated above - is a reasonable secondary motivation, not a promised outcome. What to skip: anything sold specifically as a guaranteed "AI citation schema" fix. Google doesn't require special AI markup, and the strongest available evidence - a real, sizeable 2026 study, not an absence of one - doesn't show that generic schema produces a broad citation lift for pages that are already being cited, and shows a small decline on Google's AI Overviews specifically.

← Back to the AEO fundamentals pillar
How AI engines actually choose what to cite →
Writing claims AI engines actually trust →

---

Sources: Google's statement that structured data and special schema.org markup are not required for generative AI search is quoted directly from Google's AI optimization guidance - the same verbatim quote already used elsewhere in this cluster, scoped here explicitly to what it does and doesn't claim. What structured data does for rich results and entity understanding is documented in Google's introduction to structured data and Google's Article structured-data documentation - both verified directly at the primary source. The direct evidence on schema and AI citation is from Ahrefs' matched difference-in-differences study (1,885 pages, published May 11, 2026, verified directly at the primary source) and, as a much more heavily qualified secondary source, Kurt Fischman's preprint (730 citations, 75 queries, published Feb 12, 2026, not peer-reviewed, verified directly - the author's commercial interest as founder of an AI-agent-operations company is disclosed above rather than left implicit). Page-experience and Core Web Vitals guidance, including the explicit statement that good scores don't guarantee rankings, is from Google's page experience documentation and Google's Core Web Vitals documentation - verified directly. The fetch-reliability finding is reported by Mike King at iPullRank, citing Profound's analysis of roughly 700,000 pages (published May 8, 2026, verified directly) - flagged here as preliminary industry data, not an independently published or peer-reviewed study. Crawler access context is from OpenAI's crawler overview and Perplexity's crawler documentation, already cited elsewhere in this cluster - used here only for crawler access and control claims, not extended to cover rendering, timeout, or Core Web Vitals behavior they don't document. FAQ rich results, referenced in an earlier draft of this page as a current example, were restricted to authoritative government and health sites as of September 2023 and were removed from Google Search entirely as of roughly May 2026 - confirmed via independent industry reporting - which is why that example has been replaced above.