FAQPage schema that AI actually quotes

Answer engines reward content that is already shaped like an answer. FAQPage structured data is the cleanest way to do that: a question, and a direct answer, in a format the crawler can lift without guessing.

FAQPage is a Schema.org type, published as a JSON-LD block on the page, that lists question-and-answer pairs matching what the visitor can see. To a human it changes nothing. To a crawler it turns your page from prose-that-must-be-interpreted into answers-that-can-be-copied, and for retrieval-based engines that difference decides what gets quoted.

The shape that works

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "What is answer engine optimisation?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "Answer Engine Optimisation (AEO) is the practice of improving how often and how favourably AI engines like ChatGPT and Perplexity mention a brand in their answers."
    }
  }]
}
</script>

One FAQPage block per page, containing every visible Q&A pair on it. The question text should match a real prompt word for word where possible, and the answer text must be identical in meaning to what the page visibly shows — invisible or contradicting schema is the one reliable way to get structured data ignored or penalised.

Lead with the answer

The single biggest mistake is burying the answer in a paragraph of setup. State it in the first sentence, name the alternatives, then add nuance. A model quoting you wants the conclusion, not the throat-clearing before it.

A useful length target is 40 to 70 words per answer: long enough to be complete on its own, short enough to be lifted whole. Every answer should survive being read with no page around it, because that is exactly how an engine will use it.

Keep one question per entity

Match real prompts to your questions one to one. If buyers ask "what are the best alternatives to X", that exact phrasing should be a question in your schema, answered in plain language a model can quote verbatim.

The mistakes that get schema ignored

• Schema that does not match the visible page. Crawlers cross-check; mismatches burn trust for the whole domain.
• Marketing copy as answers. "We are the leading platform for..." is not an answer, and models skip it.
• Thirty questions stuffed onto one page. Relevance dilutes; keep the set tight and move the rest to pages where they belong.
• The same FAQ block pasted across many pages. Duplicate answers compete with themselves; each Q&A should live on exactly one canonical page.
• Invalid JSON. One trailing comma silently kills the whole block — validate before shipping.

A note on Google rich results

Since 2023 Google shows FAQ rich results only for a narrow set of authoritative sites, so do not ship this expecting expandable snippets in the SERP. That is not the point anymore. The point is that answer engines read structured data when choosing what to quote, and a clean FAQPage block is the cheapest way to hand them your answer in your words.

Ship it, then verify

Adding schema is not the finish line. Run the prompt the schema targets and check whether the model now pulls your answer. If it does not, the answer is too long, too hedged, or buried below stronger sources. Tighten and run it again.

Questions, answered

Does FAQPage schema still matter after Google limited FAQ rich results?

Yes, but for a different reason. Google rarely shows FAQ rich results anymore, yet AI engines still read structured data when deciding what to quote. FAQPage schema is now less about SERP decoration and more about handing answer engines a clean, quotable version of your answer.

How long should an answer in FAQPage schema be?

Aim for 40 to 70 words. The answer must be complete when read on its own — engines lift it out of context — but short enough to be quoted whole. State the conclusion in the first sentence and use the rest for essential nuance only.

Can I put the same FAQ on several pages?

Avoid it. Duplicated Q&A pairs compete with themselves and dilute which page an engine treats as the source. Give each question one canonical home on the page most relevant to it, and link there instead of copying.

How do I check my FAQPage schema is valid?

Run the page through Google's Rich Results Test or the Schema.org validator, then confirm the JSON-LD parses and every question matches visible page content. After that, verify the thing that matters: run the target prompt in ChatGPT or Perplexity and see whether your answer gets used.