Citebase
Guides Pay $197

Structured data

Schema markup for AI search: what it does and doesn’t do

No AI company documents a special schema for getting into AI answers. Google says structured data is not required for its AI features and that there is no special schema.org markup to add. Markup is still worth having for what it already does in Search, as long as it matches the page. It is not a lever that gets you cited.

What Google says

Google is the only one of the big AI search products that writes about structured data and AI answers directly.

  • Its optimization guide for generative AI features says structured data is not required for generative AI search, and that there is no special schema.org markup you need to add. It says structured data is still worth using to be eligible for rich results.
  • Its AI features page lists, among ordinary good practices, making sure structured data matches the visible text on the page.
  • The introduction to structured data describes it as a standardized format for giving information about a page and classifying its content. Google recommends JSON-LD where your setup allows it, and says JSON-LD, Microdata, and RDFa are equally fine if the markup is valid. Most Search structured data uses the schema.org vocabulary, but Google’s own documentation is what decides how Search behaves, and schema.org has many properties Google does not use.

The general structured data guidelines set the limits. Google does not guarantee that structured data will show up in results. Do not mark up content readers cannot see. The markup must be a true representation of the page. Misleading markup can lead to a manual action that removes rich result eligibility.

What the other companies say

Very little, and it is worth being precise about that.

  • OpenAI. Its shopping help article says ChatGPT weighs structured metadata from first- and third-party providers when it picks products. That describes product data from providers and merchant feeds. It does not say ChatGPT reads schema.org markup on your pages. The publisher FAQ does not mention structured data for search.
  • Perplexity. Its crawler documentation covers robots.txt, IP ranges, and firewalls. It does not mention structured data.
  • Anthropic. Its crawler article covers ClaudeBot, Claude-User, and Claude-SearchBot. It does not mention structured data.
  • Microsoft. In its AI Performance announcement, the Bing team says clear headings, tables, and FAQ sections help surface key information. That is about the visible structure of the page, not markup.

So if someone tells you a particular schema type gets you into ChatGPT or Perplexity answers, ask where that is documented. As of this writing, it is not.

Why people believe markup helps anyway

The belief is not silly. Structured data states facts in a fixed format: this is the product name, this is the price, this is the organization behind the site. A system that reads it could use those facts. Whether a given AI product does, for a given answer, is not something any of these companies has published. Treat claims that it does as observed practice at best, and check whether the person making the claim can show their method.

What is documented is the reverse risk. Markup that disagrees with the page is a problem in Google’s guidelines. A price in JSON-LD that does not match the price on the page is the kind of conflict that helps nobody.

Markup worth having

These earn their place through what Google documents for ordinary Search. If AI products also use them, that is a bonus, not the reason.

Common types for a small product site, and what Google’s documentation says each is for.
Type What Google says it does
Organization On the homepage, helps Google understand your organization’s details and tell it apart from others, and can influence which logo is shown. The Organization docs say there are no required properties. Recommended ones include name, url, logo, sameAs, address, and contactPoint.
Product with offers The product docs separate product snippets, for pages where people cannot buy directly, from merchant listings, for pages where they can. Both can show details such as price, availability, and ratings in results.
Review and rating The review snippet guidelines say that if an organization controls the reviews about itself, its pages using LocalBusiness or Organization markup are not eligible for star reviews. Ratings must come directly from users.

A minimal Organization block for a homepage looks like this. It is an illustration of the shape, with placeholder values:

<script type="application/ld+json">
        {
          "@context": "https://schema.org",
          "@type": "Organization",
          "name": "Your Product",
          "url": "https://example.com/",
          "logo": "https://example.com/logo.png",
          "sameAs": ["https://www.linkedin.com/company/your-product"]
        }
        </script>

Every value in it should already be visible or linked on the site. The sameAs links should point to profiles you actually control.

Where product data does count

For stores, the documented link between structured product data and AI answers runs through feeds more than markup. Google’s optimization guide says Merchant Center feeds help products appear in AI responses and other Search results. OpenAI’s shopping article says ChatGPT weighs structured metadata from first- and third-party providers, and that merchants can apply to send a product feed. Product markup on the page still matters for Google’s merchant listings. It should carry the same price and availability as the feed.

FAQ markup changed in 2026

A lot of AI visibility advice still says to add FAQPage markup. In Google Search, that rich result is gone. Google’s documentation updates page says the FAQ rich result would no longer appear in Google Search starting 7 May 2026, and on 15 June 2026 Google removed the FAQ documentation because the feature is no longer shown.

That does not make a FAQ section on the page useless. Visible questions and answers help readers, and the Bing team’s note about FAQ sections is about visible structure. It means FAQPage markup no longer buys you a Google rich result, and there is no documented AI benefit to replace it. If your FAQ markup is accurate, there is no reason in Google’s docs to rush to remove it. If you were planning to add it for AI answers, there is no documented reason to.

The same updates page lists other types Google has stopped showing since 2025, including practice problems, datasets, course info, and vehicle listings. Check that page before you build a plan around any rich result.

Mistakes to avoid

  • Markup for things the page does not show. Google’s guidelines say not to mark up content readers cannot see.
  • Ratings you wrote or curated yourself. Google says reviews or ratings not by actual users may result in a manual action. In the US, the FTC’s rule on consumer reviews and testimonials separately prohibits fake reviews and undisclosed insider reviews.
  • Prices that drift. If the JSON-LD comes from one system and the visible price from another, they will disagree eventually. Generate both from the same source.
  • “AI schema” plugins. Google says there is no special markup for its AI features. A plugin that adds invented types for AI does nothing documented.
  • Treating markup as a substitute for text. Google’s AI features page says to keep important content in text. The plain sentence that says what you sell does more than any property.

How to check what you have

  1. Run your key pages through Google’s Rich Results Test, which the structured data introduction recommends for validation.
  2. View the page source and read the JSON-LD next to the visible page. Compare names, prices, and dates line by line.
  3. Remove anything that describes content no longer on the page.
  4. Then spend the time you saved on the visible copy, and on the manual check of what the answers actually say.

Related: AI search vs SEO, and llms.txt and the crawler controls.

Sources