FAQPage schema: the fastest GEO win in 2026 (and what changed)

FAQPage schema is the fastest structural win in GEO, and almost nobody sells it honestly any more. The rich result it used to earn in Google is effectively gone. The reason to add it in 2026 is different, and better: it forces you to write the exact question-and-answer pairs that generative engines lift into answers, and it labels them in a format a machine cannot misread.

The part most articles still get wrong

The part most articles still get wrong is that Google changed the rules: on August 8, 2023, Google announced it was reducing the visibility of FAQ rich results, limiting them to well-known authoritative government and health websites, and restricting How-To rich results to desktop. Search Engine Land and Search Engine Journal covered it at the time.

So if you are reading a guide that promises FAQ markup will expand your Google listing, it is out of date by three years. That has not stopped agencies selling it.

The rich result is gone. The machine readability is not, and machine readability is now the thing that pays.

Generative engines assemble answers out of passages. A question-and-answer pair is the single most convenient shape a passage can have: the question states the query it answers, and the answer is already self-contained. FAQPage markup wraps that pair in an explicit label. You are handing a retrieval system pre-cut material.

Why it is the fastest win

FAQPage schema is the fastest win for three reasons, in order of importance.

• It is an afternoon, not a quarter. Eight questions, eight answers, one JSON-LD block. Compare that with earning press mentions or rebuilding your rendering pipeline.
• It fixes the content, not just the markup. Writing eight real questions forces you to state your price, your turnaround, your coverage area, and who you are not for. Most sites discover they never answered those anywhere.
• It is directly scored. In the v6.0 GeoTest rubric, the Q&A block is 4 points, question depth is 2 and FAQPage JSON-LD backed by visible text is 2, inside the 22-point Answer Structure category. The questions you add usually lift Citation Worthiness as well, because writing them forces the numbers onto the page.

How to choose the eight questions

Choose the questions by collecting them, not by inventing them.

1. Read your inbox. The last twenty pre-sale emails contain your real FAQ. Take the questions verbatim, including the awkward ones.
2. Ask sales or support. "What do people ask you before they buy?" produces a better list than any keyword tool.
3. Check the objections. Price, timeline, what happens if it goes wrong, what is not included, who it is not suitable for. Answering these plainly is a competitive advantage precisely because competitors dodge them.
4. Look at what engines already answer. Ask ChatGPT or Perplexity your category question and see which sub-questions it raises. Those are the ones being retrieved.

Keep the phrasing conversational, because queries are conversational. "How much does a GEO audit cost?" beats "Pricing information."

How to write the answers

Four rules that decide whether an answer is quotable:

• Answer in the first sentence. No preamble. The rest of the paragraph adds nuance.
• Make it self-contained. The answer must make sense with the question removed and no surrounding context.
• 40 to 90 words. Long enough to be complete, short enough to lift whole.
• Be specific, or say why you cannot. "From $49 a month, with a free plan to start" is an answer. "It depends on your needs" is not, and if it genuinely depends, state what it depends on.

The markup

The markup is one JSON-LD block in the page head or body. Every question and answer here must also be visible in the page's text.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "@id": "https://example.com/pricing#faq",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How much does a GEO audit cost?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "A quick scan is free and covers one page. Full site analyses start on the Solo plan at $49 a month and cover up to 200 pages, with the full report usually ready in a few minutes."
      }
    },
    {
      "@type": "Question",
      "name": "How long does the audit take?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "A quick scan takes seconds; a site analysis usually finishes in a few minutes. Larger sites run longer because every page is fetched and scored individually."
      }
    }
  ]
}
</script>

Two details worth getting right. Give the node a stable @id so other schema can reference it and so you do not end up with duplicate anonymous nodes. And keep the text value identical in substance to the visible answer; do not write a keyword-stuffed version for the markup.

Five mistakes we see constantly

Five FAQ markup mistakes, and why each one hurts:

What to expect, honestly

Expect no overnight change: adding FAQ markup will not by itself put you in ChatGPT's answer tomorrow. What it does is make a page unambiguous and easy to extract from, which is a precondition, not a guarantee. In the GeoTest rubric it is a set of scored checks you can complete in one sitting, and it is a rare case where the mechanical work and the editorial work are the same work.

If you only do one structural thing this month, do this one. Then check the raw server response, because markup that arrives after JavaScript may as well not exist.

Frequently asked questions

Short answers to common questions about FAQPage schema in 2026.

Does FAQPage schema still work in 2026?

In Google's blue-link results, the FAQ rich results that FAQPage schema used to earn are largely gone: in August 2023 Google announced it was reducing FAQ rich results to well-known authoritative government and health sites. The markup itself still validates and is still parsed. Its value today is machine readability for AI answering systems, not a bigger listing in Google.

Will adding FAQ schema get me a rich result in Google?

Adding FAQ schema will almost certainly not get you a rich result in Google, unless you are a government or health authority. Treat any tool or agency that promises FAQ rich snippets in 2026 with suspicion. Add FAQPage markup for machine clarity, and for the on-page structure it forces you to write.

How many questions should a FAQPage block contain?

A FAQPage block should contain enough questions to cover the real decision, usually six to twelve. Every question should be one a customer actually asks before buying, and every answer should be complete enough to be lifted on its own, typically 40 to 90 words.

Do the questions and answers need to be visible on the page?

Yes, the questions and answers need to be visible on the page. Schema must describe content that is present and visible to a user on the same page. Markup that describes hidden or absent content is a spam pattern and can be ignored or penalised, and it also defeats the point: the visible text is what a model quotes.

Can I put FAQPage schema on every page of my site?

FAQPage schema can go on every page, but do not paste the same block everywhere. Each page should carry the questions that belong to that page's topic. Duplicated FAQ blocks across dozens of URLs add noise rather than clarity.

What is the difference between FAQPage and QAPage?

FAQPage is for questions the publisher wrote and answered. QAPage is for a page where users ask a question and other users answer it, such as a forum thread. Using QAPage for your own marketing FAQ is incorrect markup.

How much is FAQ coverage worth in the GeoTest score?

FAQ work sits inside Answer Structure, which is 22 of the 100 points in the v6.0 page rubric. The Q&A block itself is 4 points, question depth on the page is 2 and FAQPage JSON-LD backed by visible text is 2, so the markup alone is never the whole prize.

Free GEO resources: The GeoTest Benchmark Report 2026 and the AI Search Readiness Checklist