What is FAQPage schema and why does it matter for AI search?

So what is it, exactly?

FAQPage is the Schema.org structured-data type that wraps a question-and-answer block on a page. It tells search engines and AI answer engines two specific things — this content is a frequently-asked question, and this content is the answer. Cleanly separated. Quotable as a unit.

Why does that matter? Because the engine doesn't have to guess. It extracts the Q&A pair as a standalone unit instead of summarizing surrounding prose.

This is the single highest-leverage schema type for Generative Engine Optimization right now. Three reasons. The AI answer engines extract Question/Answer pairs from FAQPage JSON-LD verbatim rather than synthesizing. The schema pairs naturally with the way users ask conversational queries. And Google AI Overviews and Bing Copilot both surface FAQ-shaped content at materially higher rates than equivalent prose.

If a page on your site is shaped like a question-and-answer block, the question isn't whether you have FAQPage schema on it. It's whether you've gotten around to deploying it yet.

What does the schema actually look like?

A minimal valid FAQPage JSON-LD block:

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How fast does a Doxia Axis audit ship?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "The free AI Visibility Audit ships in five business days. The deliverable is a 14-page dossier with revenue-quantified findings."
      }
    },
    {
      "@type": "Question",
      "name": "Is the audit really free?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Yes. We deliver it free because the audit is the proof artifact. If the dossier doesn't earn the next conversation, nothing we say will."
      }
    }
  ]
}

Two extensions that improve the schema in practice.

1. speakable for voice-assistant citations

"speakable": {
  "@type": "SpeakableSpecification",
  "cssSelector": ["#faq-heading", "[data-faq-answer]"]
}

The speakable block tells voice-assistant systems which CSS selectors mark the question-heading and the answer text on the rendered page. Pair it with corresponding id="faq-heading" and data-faq-answer attributes on the actual HTML elements. We deploy this on every FAQ surface on Doxia Axis — homepage, /pricing, /services.

2. author on each Answer

For pages where the answers are operator-grade and you want the citation chain to credit the author:

"acceptedAnswer": {
  "@type": "Answer",
  "text": "...",
  "author": { "@id": "https://doxiaaxis.com/#founder" }
}

The @id link assumes you have a Person schema block elsewhere on the site with that @id. Useful for thought-leadership FAQ surfaces.

Why is FAQPage the most-cited schema type?

Three structural reasons. Worth unpacking.

Reason 1 — the engines extract verbatim

When a user asks ChatGPT "how fast does an AI audit ship?", the engine matches the question against indexed FAQPage.mainEntity[].name fields directly. If a Question matches the user's intent closely enough, the engine quotes the corresponding Answer.text verbatim.

There is no synthesis layer between the page and the response. None. This is the highest-fidelity citation path the schema vocabulary offers.

Reason 2 — the shape matches conversational queries

Users don't ask AI engines questions in keyword shape ("audit timeline doxia axis"). They ask in natural-language question shape ("how fast does a Doxia Axis audit ship?"). Pages whose Question text matches the natural-language shape get matched at higher rates than pages where the equivalent content lives in prose.

The implication for content shape — write FAQPage questions in the literal shape users would speak them. Not "Audit Timeline" as an H2. "How fast does an audit ship?" as a Question.name. The closer the schema text is to user-speech, the higher the citation rate.

Reason 3 — the engines can validate the answer is bounded

A FAQPage.Answer.text field is structurally a single answer to a single question. The engine has confidence the content represents an authoritative response, scoped to that question, with the page taking responsibility for accuracy.

This is qualitatively different from extracting a sentence from flowing prose, where the engine has to guess which surrounding context belongs.

So where does FAQPage actually belong?

Five page types where FAQPage is high-leverage.

Pricing pages. "How much does X cost?" "Is there a free tier?" "What does the free tier include?" We deploy this on /pricing with five Q&As.

Services and product pages. "What does Doxia Axis build?" "What are the diagnostic suites?" "Which tier do I need?" Live on /services.

Practice-area pages for service businesses. For a law firm — "How long do I have to file a personal injury claim in Georgia?" "What is the average settlement for a motorcycle accident?" The single highest-leverage schema deployment for legal-vertical GEO. Detail in how law firms appear in ChatGPT and Perplexity.

Product/SKU pages for e-commerce. "Is this product machine washable?" "What is the return policy?" "How long is shipping?"

Property pages for hospitality. "Is the hotel pet-friendly?" "Is breakfast included?" "What are check-in times?" The Carolina inn audit found the entire FAQ corpus was sitting in an email autoresponder rather than on the property's own site marked up as FAQPage.

Where doesn't FAQPage belong? Editorial articles (use BlogPosting or NewsArticle with mentions). Tutorials (use HowTo). Product comparisons (use QAPage with one comparative question — see the discussion at GEO vs SEO for the distinction).

What are the most common deployment mistakes?

Five we see in audits more often than any others.

1. The answers in the schema don't match the answers on the page. Schema validation tools accept this. The AI engines penalize it. The schema must reflect what is actually rendered.

2. The Question.name is keyword-shaped instead of question-shaped. "Pricing" fails. "How much does X cost?" succeeds.

3. Answers are too short to be useful. Three-word answers don't earn citation. The minimum useful answer is roughly 30 words. Our homepage FAQ averages 60 to 90 words per answer.

4. Multiple FAQPage blocks on the same page. Use one FAQPage with all questions in mainEntity[]. Not multiple separate blocks. The engines deduplicate but the structural validation can fail.

5. No corresponding visible content. Schema with no on-page rendering isn't just a Google guideline violation — it actively harms citation. The engines validate against rendered content. Hidden FAQ schema is a signal of low quality.

Want to see it deployed live?

Three pages on this site deploy FAQPage schema at production scale.

Homepage — five Q&As covering speed, differentiation, expertise, data security, qualification. Rendered as an interactive accordion at the bottom of the page. Every answer paragraph carries data-faq-answer for the speakable selector.

/pricing — five Q&As on cost, audit inclusion, tier differentiation, why-cheaper, the outcome guarantee.

/services — five Q&As on what we build, the diagnostic suites, tier mapping, custom solutions, ICP.

View source on any of the three. The JSON-LD blocks are visible in the page HTML, scoped to <script type="application/ld+json">. The corresponding visible content matches the schema text. The speakable selector points at real DOM IDs.

Where to go next

• The broader schema set: what schema matters for AI visibility.
• The audit methodology that finds FAQ-schema gaps: what is an AI visibility audit.
• A vertical example: how law firms appear in ChatGPT and Perplexity, where FAQPage was the highest-leverage gap.
• Or just request the audit: /audit. Five-business-day deliverable. Every missing schema type named. Every FAQPage deployment recommendation drafted.