Structured data for AI search: what helps, what doesn't, and what changed in 2026

Quick summary

• Google says there's no special schema.org markup for appearing in AI Overviews or AI Mode, and that structured data isn't required for its AI features.
• It does recommend continuing to use it for rich results, and that it matches the visible text on the page.
• FAQ rich results are no longer shown in Google as of 7 May 2026, according to its changelog. FAQPage markup is still valid schema.org, but it no longer gets you that result.
• OpenAI, Perplexity and Anthropic don't document how (or whether) they use structured data to choose sources. Any confident claim about it is a guess.

What it is, in two paragraphs

Structured data describes a page's content in a format machines understand without interpretation: "this is an article, written by this person, published on this date", "this is a product, it costs this much, it's in stock". The most widely used vocabulary is schema.org and the most common format is JSON-LD: a <script type="application/ld+json"> block in the HTML.

Google supports JSON-LD, Microdata and RDFa, and its introduction to structured data generally recommends JSON-LD as the easiest to implement and maintain.

What Google says about structured data and AI

Google has said it in three places, always along the same lines:

• In AI features and your website: you don't need new markup to appear in AI Overviews or AI Mode, and there's no special schema.org to add. Among best practices, it does list making sure structured data matches the visible text.
• In its generative AI optimization guide, in the mythbusting section: "overfocusing on structured data" is on the list of things you can ignore. It's not required for generative AI search, but it's worth keeping because it helps with eligibility for rich results.
• On its Search Central blog in May 2025: if you use structured data, follow the guidelines, make sure everything you mark up is visible on the page, and validate the markup.

Put simply: structured data isn't a lever for Google's AI. It's a way of accurately describing what's already there, which Google's systems take into account.

The end of FAQ rich results

For years, marking up FAQs with FAQPage was everyone's favourite trick: Google showed expandable questions under your result and you took up more space. In August 2023 Google restricted it to well-known, authoritative government and health sites. And according to its changelog:

• On 8 May 2026 it added a deprecation notice: the feature stops appearing in Search starting 7 May 2026.
• On 15 June 2026 it removed the FAQ rich result documentation.

Should you remove FAQPage markup? There's no need. It's still a valid schema.org type, accurately describes a Q&A section, and there's no sign it does harm. It just no longer earns a rich result in Google. If you keep it, make sure the page really has that section visible.

What about the other engines?

There's more noise than data here. Neither OpenAI, Perplexity nor Anthropic publishes that it uses schema.org to decide which pages to cite. Their public crawler documentation covers robots.txt, user-agents and IP ranges.

What can honestly be said:

• A crawler reading your HTML sees the JSON-LD, like the rest of the page. If the markup clarifies who you are, what you sell and what it costs, it doesn't get in the way.
• What does weigh in every engine is the visible text. If a fact only exists in JSON-LD and not on the page, don't expect it to be cited.
• Be wary of anyone selling "schema for ChatGPT" as a guarantee. No public documentation supports it.

Markup that's still worth it

Roughly from most to least useful for a typical company site:

Organization

On the homepage or About page: name, logo, URL, official profiles (sameAs), contact details. It helps search engines connect your brand to your site and profiles. For businesses with a physical location, LocalBusiness with address, hours and phone, consistent with your Google Business Profile.

Article or BlogPosting

On articles: headline, author, publish and update dates (datePublished, dateModified). Dates are particularly useful on content that goes stale.

Product and Offer

On product pages: price, currency, availability, ratings if they're real and visible. Google regularly extends what it supports here (its 2026 changelog includes updates on categories and sale prices).

BreadcrumbList

Breadcrumbs describe site structure. They're cheap and Google uses them to show the path in results.

FAQPage

See above: valid, but no rich result in Google any more. Keep it only if the section exists and is useful to readers.

A minimal Organization example

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Acme Booking",
  "url": "https://acme.example",
  "logo": "https://acme.example/logo.png",
  "sameAs": ["https://www.linkedin.com/company/acme-booking"]
}

Short, accurate and consistent with the page is better than long and speculative. Only add properties you can keep up to date.

Mistakes we see often

1. Marking up content that isn't visible. Google's general structured data guidelines explicitly forbid it: don't mark up content that's not visible to readers.
2. JSON-LD that doesn't parse. One stray comma and the whole block is ignored. It happens a lot when it's pasted by hand into the custom code field in Framer or Webflow.
3. Stale data. The JSON-LD says €29 and the page says €39. Generate the markup from your CMS and it can't drift.
4. Made-up ratings. In July 2026 Google added a guideline on fake and undisclosed incentivized reviews. It's also the fastest way to lose a reader's trust.
5. Several conflicting Organization entities across pages (different names, different logos).

How to check

• Google's Rich Results Test and the schema.org validator, to see whether the markup parses and what it detects.
• Search Console enhancement reports, for errors at scale.
• Citable's weekly audit checks the JSON-LD blocks on your pages and flags any that don't parse, among other technical checks. Articles written by Citable ship with Article and FAQPage JSON-LD generated from the article itself, so the markup always matches what's visible. See what each plan includes on the pricing page.

FAQ

Does structured data help you appear in ChatGPT?

OpenAI doesn't document using it to choose sources. Its crawler sees JSON-LD like the rest of the HTML, but what gets cited is visible text. Use it to describe your content accurately, not as a ChatGPT lever.

Do I need schema.org to appear in AI Overviews?

No. Google says structured data isn't required for its AI features and there's no special markup to add. It's still useful for rich result eligibility.

Should I remove FAQPage markup now that Google doesn't show FAQ rich results?

There's no need. FAQPage is still a valid schema.org type and correctly describes an FAQ section. The only change is that, since 7 May 2026, Google no longer shows it as a rich result.

JSON-LD, Microdata or RDFa?

Google supports all three and generally recommends JSON-LD as the easiest to implement and maintain.

Conclusion

Structured data isn't a shortcut to AI: Google says so plainly and the other engines haven't said otherwise. It's still useful for describing precisely who you are and what you offer, and for the rich results that remain. Mark things up properly, keep it in line with what's visible, and put the real effort into the content that gets cited.