Structured data for Google and AI search: what is documented and what is not

What is structured data?

Structured data is a machine-readable description of page content. It tells a search engine explicitly: this page is an article, written by this person, published on this date. The vocabulary comes from Schema.org. Google, Bing and Yahoo launched the project in June 2011, and Yandex joined later.

Schema.org is extensive. According to Schema.org, release 30.1 from September 2026 contains 826 types and 1,540 properties. Google uses only a small part of them for search features.

Google accepts three formats: JSON-LD, Microdata and RDFa. In its introduction to structured data, Google recommends JSON-LD in most cases because it is the easiest to implement and maintain. JSON-LD sits in its own script block in the HTML and does not need to be woven into the visible text.

The example shows the basic structure. @context points to the vocabulary, @type sets the type. Nested objects describe further entities such as the author and the publisher. Each property corresponds to a field in the Schema.org vocabulary.

What does Google officially say?

Google describes structured data as an aid to understanding. According to Google, it provides explicit clues about the meaning of a page and can enable more engaging search results. Rich results are not guaranteed. Three rules matter most:

• Visibility: markup may only describe content that users can see on the page. This is set out in the general guidelines.

• Penalties: for spammy markup, Google can issue a manual action. The page then loses eligibility for rich results.

• Testing: Google points to the Rich Results Test and the URL Inspection tool in Search Console.

For AI features, the position is clear. In its documentation on AI features, Google writes that there is no special Schema.org markup you need to add for AI Overviews or AI Mode. Existing markup should match the visible text. More on the format in our article on AI Overviews.

Which rich results remain?

Google has cut the number of rich results considerably since 2023. In August 2023, Google limited FAQ rich results to well-known government and health websites. Google retired HowTo rich results completely in September 2023. In June 2025, Google announced it would phase out seven more types, including Course Info, Claim Review and Vehicle Listing.

According to Google, this does not affect rankings. Unused markup does no harm either. You do not have to remove FAQ markup, but it is no longer worth investing effort in new FAQ snippets.

Which types are worth using?

Worthwhile markup clearly describes your website and its content. The table classifies common types. The AI search column is an editorial assessment, not a vendor statement.

Author details deserve care. Person markup that points to an author page supports what Google describes as E-E-A-T. Google does not confirm any ranking effect from it.

What is documented for AI systems?

Little is documented, much is claimed. The only publicly reported statement from a major provider comes from Microsoft. According to Search Engine Roundtable, Fabrice Canel, Principal Product Manager at Microsoft Bing, confirmed at SMX Munich in March 2025 that schema markup helps Microsoft's LLMs understand content. There is no official Bing documentation on this.

OpenAI, Perplexity and Anthropic have published no statements on structured data. Whether their systems evaluate JSON-LD when they fetch a page remains open. Whatever the markup says should therefore always appear in the visible text too.

Which mistakes should you avoid?

Most problems arise when markup and page drift apart. These mistakes are common in practice:

• Marking up hidden content: reviews or questions in the markup that do not appear on the page violate Google's guidelines.

• Outdated values: prices or dates in the markup differ from the visible text.

• Wrong type: a category page listing many products carries the markup of a single product page.

• Duplicate entities: several Organization blocks with different names describe the same company.

• Unchecked plugins: CMS extensions generate default markup that does not fit the page.

One rule sums it up: markup repeats what the page says. It adds no statements that readers cannot see. Following this rule meets the most important requirement in Google's guidelines.

A practical approach

1. Audit: test key page types with the Rich Results Test and the Schema Markup Validator.

2. Set the basics: Organization on the home page, Article with Person on editorial pages, BreadcrumbList on all subpages.

3. Add specific types: use Product, LocalBusiness or Event only where the page actually shows these things.

4. Link entities: use sameAs to point to official profiles, for example on Wikidata or LinkedIn.

5. Plan maintenance: prices, dates and author details in the markup change along with the content.

Structured data is part of the technical foundation that SEO and GEO share. How the two disciplines connect is explained in our article GEO vs. SEO. Start by auditing your five most important page types.