The Best Markup for FAQ Sections AI Engines Actually Cite

Google just finished burying FAQ rich results. They stopped showing in search on May 7, 2026, and by June the rich-result report and the Rich Results Test support are gone too. If you read that as "FAQ schema is dead," you are about to make an expensive mistake.

I suggest to keep the FAQPage markup. The search feature died; the markup did not. AI answer engines still parse it to pull and attribute answers, and a clean FAQ block is one of the easiest things on your page for a model to quote.

The reason is structural. Each question and answer is a self-contained, quotable unit.

“

"Our findings provide empirical support that adding Schema.org metadata – particularly FAQ schemas that supply question-answer pairs – is associated with significantly higher chances of being recognized or cited by AI chat systems like ChatGPT."

The model does not have to untangle a paragraph to find the claim - the claim is already isolated. That is exactly the shape an AI answer wants to lift.

But "having an FAQ" is not the same as having FAQ markup an engine will cite. Here is what actually works.

The pattern that gets cited

The best markup for FAQ sections is FAQPage JSON-LD that mirrors visible Q and A on the page, where every answer is self-contained. Three properties, in order of importance:

1. Self-contained answers. Each answer must make sense lifted out of context. "It depends on your plan" is unquotable. "Free plans include 3 audits per month; paid plans are unlimited" is a citation.

“

A well-written FAQ section with correct schema markup is one of the most effective ways to get your content cited by answer engines.

1. Visible parity. The Q and A in your JSON-LD must match what a human sees on the page. Markup that describes content the user cannot see is the fastest way to get ignored or penalized.
2. Valid FAQPage JSON-LD. Use the schema.org FAQPage type with mainEntity Question and acceptedAnswer. Plain Q and A headings still help, but the structured version is what makes extraction reliable.

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "Your exact visible question?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "A self-contained answer with the specific number or fact."
    }
  }]
}

Three mistakes that get you skipped

• Vague answers. "It varies," "contact us," "many factors apply." Nothing to quote, nothing gets cited.
• Markup that does not match the page. JSON-LD questions the visitor cannot find. Engines cross-check, and a mismatch costs you trust.
• One giant answer. Cramming five sub-points into one answer defeats the self-contained unit. One question, one clean answer.

Google killed the rich result. Keep the markup.

The timeline is the whole argument, so here it is. In August 2023 Google limited FAQ rich results to well-known government and health sites. Then it removed them outright: gone from search on May 7, 2026, with the rich-result report and Rich Results Test support retired in June 2026, and Search Console API support ending in August.

That is a display change, not a content one. Google's own FAQPage documentation says the structured data can stay in place - you do not need to rip it out. And the engines that increasingly decide whether you get mentioned at all - ChatGPT, Perplexity, Google's AI surfaces - still read it to extract and attribute answers.

So the people deleting their FAQ markup in 2026 are optimizing for a search feature that no longer exists, and throwing away citation signal that does. From auditing 50+ brands, the FAQ block is consistently one of the highest-yield sections for getting quoted. Do not gut it because a blue-link snippet went away.

Validate it - just not with the old tool

One thing did change for the worse. The Rich Results Test stops checking FAQPage in June 2026, so the validator half the internet still recommends is now the wrong one.

Use the Schema.org validator instead. It parses FAQPage, flags malformed blocks and missing properties, and does not depend on Google keeping a rich result alive. Re-validate every time you change a question - one broken property can drop the whole block.

GEOlikeaPro's FAQ Generator builds citation-ready FAQ sections with valid FAQPage markup in one click - self-contained answers, visible parity, clean JSON-LD. Try it free and see which of your answers AI engines can actually quote.