Which Schema Markup Helps AI Engines Cite Your Site? AEO vs SEO Schema

Does schema markup help AI engines cite your site?

No schema type has been shown to cause AI citations. The only controlled test we found, run by Ahrefs, saw citations stay flat or dip after pages added JSON-LD.

Ahrefs' study of schema and AI citations tracked 1,885 pages that added JSON-LD between August 2025 and March 2026, matched against 4,000 control pages. It compared citations in the 30 days before and after the markup went live.

• AI Overviews fell. Citations dropped 4.6%, which Ahrefs calls a small but statistically significant decline against the matched controls.
• AI Mode barely moved. Its 2.4% rise was statistically indistinguishable from zero.
• ChatGPT barely moved either. Its 2.2% rise was also indistinguishable from zero.

The same study found AI-cited pages were far more likely to carry JSON-LD than uncited pages. That is a correlation, and in our reading, sites that maintain schema tend to maintain everything else too.

Every page in the dataset already had 100+ AI Overview citations in February 2025, before schema was added. So the test covers pages AI already cites and says nothing about pages it ignores.

Ahrefs still sees uses for schema in rich results, voice assistants, knowledge graphs and entity recognition, but its data does not back adding it only for more AI citations on pages that are already visible.

What does Google say about schema for AI Overviews and AI Mode?

Google says its AI features need no special schema, and it still recommends structured data for rich results and for making a page's meaning explicit.

Google's AI optimization guide says "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add."

The guide still recommends structured data for SEO, because it helps pages qualify for rich results on Google Search. It also warns that many suggested AEO and GEO hacks are not effective or supported by how Google Search works.

Google's page on AI features in Search says the same for AI Overviews and AI Mode. Its introduction to structured data gives the reason to keep markup anyway: it provides explicit clues about the meaning of a page.

Google's own pipeline differs from a chatbot fetching your page. AI Overviews draw supporting results from Google's index, and AI Mode runs a query fan-out of related searches across subtopics and data sources.

Google structured data engineer Ryan Levering said at Search Central Live Toronto in April 2026 that schema is used as context served to models during fan-outs, per a write-up on suganthan.com. That is a conference remark reported second-hand, and it covers Google only.

Do ChatGPT, Claude and Perplexity read JSON-LD?

Nobody outside those companies knows for sure, and the public tests disagree. Search Engine Land reports that ChatGPT, Perplexity and others have not publicly confirmed whether they keep schema when crawling or use it to extract facts.

• Visible HTML only. Per suganthan.com, a searchVIU experiment in October 2025 tested 5 systems: ChatGPT, Claude, Perplexity, Gemini and Google AI Mode. As Ahrefs reports it, every system extracted only visible HTML during direct retrieval and ignored JSON-LD, hidden Microdata and hidden RDFa.
• Read as plain text. In a test covered by Search Engine Roundtable, ChatGPT and Perplexity found an address that existed only in made-up, invalid JSON-LD, and the tester concluded they read it like any other page text.
• Stripped from some training data. The cleaning rules for the C4 web dataset drop any page containing curly brackets, which removes most JSON-LD, per suganthan.com.
• Microsoft says it helps. Microsoft's advertising blog says schema helps search engines and AI systems understand your content. Search Engine Land reports that Bing's Fabrice Canel confirmed in March 2025 that schema helps Microsoft's LLMs understand content for Copilot.

In every test, the visible text got read. Put every fact you mark up in the page copy as well (see step 4 of the checklist below).

How is schema for AI search different from SEO schema?

Both use the same schema.org vocabulary but differ in goal, in what counts as done and in what breaks them, as the table below shows.

One block of markup can serve both: a Product with price and reviews can earn a rich result and state facts at once.

Which rich results has Google retired?

Google has dropped HowTo, FAQ and a batch of rarer rich result types, so markup written only for those displays no longer produces them in Google Search.

• HowTo first. Google stopped showing How-to rich results on desktop as of September 13, 2023, and called the result type deprecated.
• FAQ in two steps. Google's post on HowTo and FAQ changes limited FAQ rich results to well-known, authoritative government and health sites in August 2023. The feature stopped appearing for everyone on May 7, 2026, and Google's Search Central changelog records removing its documentation in June 2026.
• A batch of other types. Google's June 2025 simplification of search results announced the phase-out of Book Actions, Course Info, Claim Review, Estimated Salary, Learning Video, Special Announcement and Vehicle Listing.

Each notice covers how Google Search displays results. None of them says how other engines treat the same markup.

Which schema types are worth adding?

Start with the types that state identity and authorship: Organization, Person and Article, plus Product if you sell online. Everything else is optional.

What to add

• Organization with sameAs. Your company's name, logo, URL and official profiles, set out in the next section.
• Person for every author. A named writer with a profile page and real social profiles, nested inside the Article.
• Article or BlogPosting. Headline, author, publisher, and publish and update dates for each post.
• Product and Offer for stores. Price, availability and specifications that match the product page; our e-commerce AEO guide covers the store side.
• BreadcrumbList for navigation. States where a page sits in your site's hierarchy.

What to skip

• FAQPage and HowTo. Google no longer shows either display; see the retired list above.
• Anything sold as AI schema. See the Google section above.

What does good Organization and Article markup look like?

Good markup names each entity once, gives it an @id, and links it to profiles that prove who it is. The example at the end of this section, trimmed from this page's own JSON-LD, shows the pattern.

Organization with sameAs

Google's Organization markup guide says organization structured data on your home page can help Google understand your administrative details and disambiguate your organization in search results.

The Organization guide defines sameAs as the URL of a page on another site with more information about you, such as a social or review profile, and allows more than one. Google can also make general use of sameAs and other schema.org data.

Schema.org's definition of sameAs is stricter: a reference page that unambiguously indicates the item's identity, such as its Wikipedia page, Wikidata entry or official website.

A sameAs link to your Wikidata item ties your brand to an entry that feeds Google's Knowledge Graph, per suganthan.com. Our guide to Wikipedia and Wikidata for AEO covers getting that entry right.

Article with a Person author

Google's Article markup guide says to use Person for people and Organization for organizations, never Thing. It strongly recommends a type plus a url or sameAs for each author.

Keep author.name to the name alone, with no job title or publisher in it. Dates go in ISO 8601 format, ideally with a timezone.

Google recommends JSON-LD where your site's setup allows it. Microsoft notes that schema is usually added as a JSON-LD script, often through the CMS.

The markup on this page

Here is the JSON-LD on this article, trimmed to the parts that carry identity. The live version also includes datePublished, dateModified and the logo as an image object.

{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Organization",
      "@id": "https://onlyaeo.com/#organization",
      "name": "OnlyAEO",
      "url": "https://onlyaeo.com",
      "sameAs": [
        "https://www.linkedin.com/company/onlyaeo",
        "https://x.com/Only_AEO",
        "https://www.youtube.com/@OnlyAEO"
      ]
    },
    {
      "@type": "Article",
      "headline": "Which Schema Markup Helps AI Engines Cite Your Site? AEO vs SEO Schema",
      "author": {
        "@type": "Person",
        "@id": "https://onlyaeo.com/authors/tejas-patil#person",
        "name": "Tejas Patil",
        "url": "https://onlyaeo.com/authors/tejas-patil",
        "jobTitle": "Co-founder",
        "sameAs": [
          "https://www.linkedin.com/in/tejaspa",
          "https://x.com/tejastrek"
        ],
        "worksFor": { "@id": "https://onlyaeo.com/#organization" }
      },
      "publisher": { "@id": "https://onlyaeo.com/#organization" },
      "mainEntityOfPage": "https://onlyaeo.com/blog/schema-markup-for-aeo-beyond-traditional-seo-structured-data"
    }
  ]
}

The Article names the Organization as publisher by @id instead of repeating it, and the author links back through worksFor. Each sameAs URL is one of our real profiles, and the visible byline links to the same author page.

How do you check your markup is right?

Run it through a validator, then compare it with the page by eye, since no tool can tell whether your facts are true. Work through this list before a template ships:

1. Parse the JSON-LD. A missing comma or bracket breaks the whole block, so confirm it is valid JSON first.
2. Test supported types in Google's Rich Results Test. Google calls the tool an easy way to validate structured data and, in some cases, preview a feature in Search.
3. Test everything else in the Schema Markup Validator. Types with no Google rich result still need a syntax check, and schema.org's validator reads any type.
4. Match the visible text. Google's AI features page lists structured data that matches the visible text among its best practices. Its structured data policies say not to mark up content readers cannot see: if the JSON-LD describes a performer, the body must describe that same performer.
5. Confirm every sameAs URL. Each one should load and be your real profile, with no abandoned handles or look-alike accounts.
6. Check the @id links. Each Article should point to the same @id values as the example above.
7. Re-test after template changes. A theme update or CMS migration can drop or duplicate the block without anyone noticing.

Where does schema fit in an AEO plan?

Schema comes after the basics that decide whether AI engines can read and quote you at all. Work in roughly this order:

1. Crawler access. If robots.txt blocks AI crawlers, no markup helps; see the robots.txt mistake that blocks AI crawlers.
2. Content AI can lift. Answers in short passages that stand alone; see content structure that gets cited by AI.
3. Entity and mentions. One consistent brand entity, named on sites engines already trust; see building a brand entity AI engines recognize.
4. Schema as hygiene. The small set from this article, validated and kept in step with the page.
5. Measurement. A fixed set of buyer prompts, run on a schedule; see how to measure your AI visibility.

If you would rather hand the work off, OnlyAEO is a fully managed service: our team does the AI search work, and your team reviews and approves every piece before it goes live. First measurable results usually take at least 60 days, and we do not promise specific rankings or citations.

How we tested: what ChatGPT and Claude said when we asked

Before writing, we ask ChatGPT and Claude the buyer's question, read every page they cite in full, and turn each search they ran into a section of the article. We ran it on this article's main question in October, the day we rewrote it.

ChatGPT searched and cited pages from Google, schema.org and Bing. Its answer on AI schema matched Google's position above.

Its checklist then recommended ClaimReview, HowTo and FAQPage-style markup anyway, all types Google no longer shows as rich results (see the retired list above).

Claude searched for this year's studies and cited mostly AI search vendor and agency blogs. Its verdict lined up with the Ahrefs test above.

Answers vary by run and by user, so this is a snapshot. It showed us that an engine's markup advice can trail Google's documentation, so check the docs before you ship.

More questions about schema markup and AI search

Is Microdata or RDFa as good as JSON-LD?

For Google, yes: its structured data guide says all 3 formats are equally fine as long as the markup is valid and follows the feature's documentation. Bing's webmaster guidelines say its crawlers do not prefer one specification over another.

Should we remove FAQPage markup now that Google no longer shows FAQ rich results?

There is no need to. Google says structured data that is not being used does not cause problems for Search, though it also has no visible effect. Keep the visible questions and answers if your readers use them.

Does an llms.txt file replace schema markup?

Not for Google. Google says you do not need new machine-readable files, AI text files, markup or Markdown to appear in Search or its generative AI features, because Google Search does not use them.

Does Organization markup need to go on every page?

No. Google recommends placing it on your home page or on a single page that describes your organization, such as the about page. Its logo property can help Google understand which logo to show in Search results and knowledge panels.