FAQ Schema vs QA Schema for AI: What Engines Use

TL;DR

Google restricted FAQ rich results to health and government sites in August 2023, then deprecated the feature entirely on May 7, 2026.
The schema itself is still valid Schema.org vocabulary. AI engines like ChatGPT, Perplexity, and Google's AI Overviews appear to ingest it as content context, even though Google no longer renders it visually.
Use FAQPage only when the author is answering their own questions. Use QAPage for community-contributed answers (forums, Stack Overflow-style pages).
Mislabeling a forum thread as FAQPage is a documented quality violation. Mislabeling editorial Q&A as QAPage underuses the markup.
For GEO, JSON-LD in the <head> is the recommended format. LLMs treat the markup as plain text signals, not as a structured graph.

FAQ and Q&A markup were never identical, but for years SEOs treated them as interchangeable because both produced similar SERP treatments. That ambiguity ended with Google's August 2023 restriction and the May 2026 deprecation of FAQ rich results. The question now is whether the markup still matters for generative engines — and if so, which type to deploy on which page.

What Google actually changed (and what it didn't)

In August 2023, Google narrowed FAQ rich result eligibility to "well-known authoritative government and health websites," per the Search Central blog post announcing the change. For everyone else, FAQ markup stopped producing the expandable accordion in search results.

The full deprecation came in May 2026, when Google retired FAQ rich results entirely, as covered in Search Engine Journal. Critically, Google did not deprecate the schema vocabulary itself. The FAQPage documentation still exists, and the markup is still valid Schema.org. What disappeared is the visual SERP feature — not the parsing.

This is the nuance most SEO writeups miss: deprecation of a rich result is not deprecation of the underlying data. Crawlers still see it. AI engines that train on or retrieve from web content still tokenize it.

FAQPage vs QAPage: the definitional split

Schema.org draws a clear line that Google's docs reinforce:

FAQPage is for pages where the page author publishes a list of questions and authoritative answers. Think product FAQs, support docs, glossary pages. One author, one canonical answer per question.
QAPage is for pages where users submit a question and one or more users (or the community) respond with answers. Stack Overflow question threads, Quora-style pages, community forums. Multiple answers, often with vote counts and accepted-answer flags.

Neil Patel's guide reiterates Google's rule: if the same question has multiple user-submitted answers, it's QAPage. If you authored both the question and the answer, it's FAQPage. Choosing the wrong type isn't cosmetic — Google historically treated mismatches as a structured data quality violation, and that classification still applies to Search Console error reports.

FAQPage example (JSON-LD)

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "Does FAQ schema still help with AI search?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "Yes. While Google deprecated FAQ rich results in May 2026, the markup is still parsed by crawlers and appears to be consumed by generative engines as content context."
    }
  }, {
    "@type": "Question",
    "name": "Is FAQPage schema deprecated?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "The rich result is deprecated. The schema vocabulary is not — it remains valid Schema.org and Google still parses it without rendering a SERP feature."
    }
  }]
}

QAPage example (JSON-LD)

{
  "@context": "https://schema.org",
  "@type": "QAPage",
  "mainEntity": {
    "@type": "Question",
    "name": "How do I migrate from FAQPage to QAPage markup?",
    "answerCount": 2,
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "Use JSON-LD in the head, point @type to QAPage, and include answerCount plus upvoteCount where applicable.",
      "upvoteCount": 14,
      "author": { "@type": "Person", "name": "Jane Dev" }
    },
    "suggestedAnswer": [{
      "@type": "Answer",
      "text": "You don't migrate — they describe different page types. Keep FAQPage for editorial FAQs.",
      "upvoteCount": 8,
      "author": { "@type": "Person", "name": "Sam Ops" }
    }]
  }
}

Notice the structural differences: QAPage expects answerCount, supports multiple suggestedAnswer entries, and includes author and upvote metadata. FAQPage has a flat list of Question → acceptedAnswer pairs with no community signals.

Do AI engines actually use this markup?

This is where evidence gets thinner. No major LLM provider has published a definitive "we ingest FAQPage" statement. But WordStream's analysis and industry testing both converge on a practical observation: pages with well-formed schema appear more often in AI Overviews and Perplexity citations than equivalent unmarked pages.

The mechanism is mundane. LLMs don't traverse JSON-LD as a typed graph the way a knowledge engine does. They read it as text in the HTML payload. That means the question-answer pairs you put in FAQPage.mainEntity end up as clean, declarative training-like signals — short questions, direct answers, no surrounding navigation chrome. That format is exactly what retrieval pipelines reward.

So the markup helps, but not because of magic schema semantics. It helps because it forces you to write content in the exact Q&A shape generative engines prefer to cite.

Practical recommendations for 2025+

Keep FAQPage on editorial FAQs. Even without rich results, the markup creates clean Q&A blocks that AI engines consume well. Don't strip it.
Use QAPage for forums and community threads. If your platform has user-generated answers, label them correctly. Mismatched types are a known structured data violation.
Use JSON-LD, not microdata. It's cleaner to parse, easier to maintain, and is Google's stated preference.
Don't duplicate. If you mark up an FAQ section with FAQPage, make sure the visible page contains the same questions and answers. Hidden-only content violates Google's policy.
Write the answers as if they'll be quoted verbatim. Because in AI Overviews, they often are.

FAQ

Is FAQ schema worth implementing in 2025?

Yes, despite the loss of rich results. The markup is still parsed by Google and appears to influence AI engine citations. The cost of adding JSON-LD is trivial relative to the visibility upside in generative search.

Can I use FAQPage on any page with questions?

No. Google's guidelines restrict FAQPage to pages where the author provides authoritative answers to their own questions. Marketing pages with rhetorical questions, product pages, and category pages are not eligible.

What happens if I use FAQPage on a forum page?

It's a structured data quality violation. Google may ignore the markup, flag it in Search Console, or in repeated cases issue a manual action. Use QAPage for any page with user-submitted answers.

FAQ schema vs QA schema for AI: which one engines actually use