Structured data and AI search

For AI features

Google says there is “no special schema.org structured data that you need to add” to appear in AI Overviews or AI Mode. Among SEO practices that remain worthwhile, it lists making sure structured data matches the visible text on the page.

Officially documented · Google Search Central

Formats

Google supports JSON-LD, Microdata and RDFa, and says all three are “equally fine” when valid. It recommends JSON-LD where possible as the easiest to implement and maintain at scale.

Officially documented · Google Search Central

Guidelines that apply to all markup

“Don’t mark up content that is not visible to readers of the page.” Markup must describe the same thing the page shows. If the JSON-LD describes a performer, the HTML body must describe that performer.

Officially documented · Google Search Central

A structured data manual action makes a page lose eligibility for rich results; Google states it “doesn’t affect how the page ranks in Google web search”. Google also does not guarantee that correctly marked-up structured data will appear in results.

Officially documented · Google Search Central

FAQ and HowTo rich results are gone

Google deprecated the FAQ rich result in May 2026; it “will no longer appear in Google Search starting May 7, 2026”. Google removed its documentation on June 15, 2026. FAQPage markup no longer produces a rich result in Google Search.

Officially documented · Google Search Central

Google ended HowTo rich results in 2023. It first limited them to desktop in August, and as of September 13, 2023 Google Search “no longer shows How-to rich results on desktop”, which made the result type deprecated. FAQ rich results were cut back the same month to well-known government and health sites. Google added that “there’s no need to proactively remove it”: unused structured data does not cause problems for Search but has no visible effect.

Officially documented · Google Search Central Blog

Microsoft: feeds, markup and the page must agree

Microsoft’s guide for retailers on Copilot shopping describes three kinds of data: crawled web data, product feeds, and what AI agents read on the live site. Structured data is not a fourth source. The guide asks for prices and stock to match across the product feed, on-site schema and what shoppers see.

Copilot breaks a shopping query down using crawled web data and product feeds. In Microsoft’s example, a product makes the top three recommendations because feeds show a competitive price and in-stock status. The guide adds that “traditional SEO remains essential because AI systems perform real-time web searches frequently throughout the shopping journey”.

Officially documented · Microsoft Advertising

The guide’s technical foundations start with deploying Product, Offer, AggregateRating, Review, Brand, ItemList and FAQ schema types. Price and inventory should sync in real time between product feeds and on-site schema. Values should stay consistent across feed, on-site schema and user-facing displays. Finally: “Ensure rendered DOM contains the same facts consumers see — never serve different HTML to bots.”

Officially documented · Microsoft Advertising

On agents: “Without your live site working properly, the sale fails even if your feed and crawled data were perfect.” On claims, the guide warns against exaggerated or unverifiable wording, saying AI systems penalize low-trust language.

Officially documented · Microsoft Advertising

This is a marketing guide from Microsoft’s ad business, not ranking documentation. It describes Copilot’s shopping features, not AI answers in general. Its advice agrees with Google’s guidelines above: markup must describe what is visibly on the page.

What has been measured

Ahrefs compared 1,885 pages that added JSON-LD with matched pages that did not. Against the controls it found no measurable citation gain in ChatGPT (+2.2%) or Google AI Mode (+2.4%), and citations in Google AI Overviews fell 4.6%. All the pages were already heavily cited, so the study says nothing about pages that AI systems do not cite yet.

Observed in practice · observed 2026-05 · Ahrefs

Observed in 2025-10, more than 9 months ago. Answer engines change often; re-test before relying on this.

In searchVIU’s October 2025 test, none of five AI systems reported a price that existed only in JSON-LD. Hidden Microdata and RDFa were ignored as well. ChatGPT and Gemini, which fetched the page live, found prices in visible text.

Observed in practice · observed 2025-10 · searchVIU

A University of Toronto study of 158,099 AI answers (October 2026) found FAQ blocks, structured data, summary boxes and page speed effective only when different websites were compared. Comparing pages of the same website, markup was not associated with more citations. The authors say the results are associations.

Research-backed · Chen and Koudas (University of Toronto), arXiv:2610.15033

SAGEO Arena (KDD 2026), a reproducible search pipeline over a fixed corpus of 171,003 web documents, found that using structural information such as schema markup helped mitigate the loss of visibility that existing optimization approaches caused in retrieval and reranking.

Research-backed · Kim et al. (Yonsei University), arXiv:2602.12187

These are single studies, with limits listed in Does structured data change AI search citations?. The lab result (SAGEO Arena) and the field results (Ahrefs, Toronto) point in different directions, and this entry does not reconcile them.

What is not established

• That adding structured data increases citations on pages AI systems do not cite yet. Neither study above tested it. See the open research design in Does structured data change AI search citations?.
• Whether structured data is used when AI features answer from a search index. searchVIU did not test Google AI Overviews or Bing Copilot.

How this wiki uses it

Each entry outputs Article (or CollectionPage for hubs) and BreadcrumbList JSON-LD describing what is visibly on the page. It includes no rich-result types the page does not actually contain. See Article structured data for why Article rather than a subtype.