FAQPage Schema for AI | Customer Impact

Yes, FAQPage schema still makes sense, but no longer for the reason most companies originally added it. Google has removed the FAQ rich results from the search results, so that expandable block under your link is disappearing. Even so, the underlying Q&A structure remains one of the most extractable content forms for AI engines like ChatGPT, Perplexity and Google’s AI Overviews. In this article you will read exactly what has changed, why FAQ structure still works for AI, and how to implement it correctly with copy-paste JSON-LD for B2B.

What exactly changed at Google?

Since 7 May 2026, Google no longer shows FAQ rich results in the search results. That is the final chapter of a process that already began in 2023. Back then Google announced that FAQ rich results would only appear for authoritative government and health websites. The reason was simple: the FAQ block was being widely abused to grab more space in the SERP, often with questions nobody was asking.

In May 2026, Google pulled the plug for good for everyone. In June 2026, Google also removes the associated report in Search Console, the search appearance filter and the support in the Rich Results Test. The documentation page about FAQPage has meanwhile been taken down as well.

Important to know: you do not have to remove your existing FAQPage markup. Google simply ignores it for display, it causes no errors and it does not harm your SEO. Removing it is optional, not required. That distinction is crucial, because it means the schema is free to play a different role: fuel for AI.

Why does FAQ structure still work for AI engines?

FAQ structure works for AI because it ties each question to a single self-contained, bounded answer, exactly the kind of block a language model can easily pick up. An AI engine composing an answer looks for passages that stand on their own and fully answer a concrete question. A Q&A pair is the purest form of that.

But be honest about how that works technically, because that is where a lot of GEO advice goes astray. AI models do not read your JSON-LD as structured data. Controlled tests in 2026 indicate that a language model processes your schema as ordinary text, not as a separate data layer it interprets on its own. In other words: the schema itself is not a magic citation button.

The gain lies elsewhere. By implementing FAQPage schema, you force yourself to pour your content into a tight Q&A pattern: a clear question, followed by a short and complete answer. That visible pattern on the page is what AI engines extract. The schema is therefore not a cause, but a discipline that enforces the right content pattern. On top of that, clean structured data still helps Google understand your content for AI Overviews and AI Mode, even though the rich result no longer appears.

If you want to understand more deeply how AI reads and extracts your page, read our guide on content architecture for AI extraction. FAQ schema is just one tactic there within a broader whole.

How do you implement FAQPage schema?

You implement FAQPage schema with a small block of JSON-LD in the <head> or <body> of your page, containing each question as a Question and each answer as an acceptedAnswer. The golden rule: the schema must match exactly the questions and answers that are visible on the page. Hidden FAQs or answers that diverge from the visible text are a bad idea, for people and for AI.

Here is a copy-paste example for a B2B context that you adapt to your situation:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "Voor welke bedrijven is jullie oplossing geschikt?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Onze oplossing is gebouwd voor B2B-bedrijven met een complexe verkoopcyclus en meerdere beslissers. Typisch werken we met organisaties van 20 tot 500 medewerkers in dienstverlening, software en industrie."
      }
    },
    {
      "@type": "Question",
      "name": "Hoe snel zien we resultaat?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "De meeste klanten zien de eerste meetbare effecten binnen drie tot zes maanden. De exacte termijn hangt af van je uitgangspositie, je sector en de competitiviteit van je markt."
      }
    },
    {
      "@type": "Question",
      "name": "Wat kost een samenwerking?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Een traject start typisch vanaf een vast maandbedrag dat we afstemmen op je doelen en scope. Tijdens een gratis intake bepalen we samen welke aanpak past bij je budget en ambitie."
      }
    }
  ]
}
</script>

Validation is no longer done via Google’s Rich Results Test, because it no longer supports FAQ. Use a general schema validator instead, such as the Schema Markup Validator from Schema.org, to check that your JSON-LD is syntactically correct.

Which questions do you put in your FAQ for B2B?

Put in your FAQ the questions your buyers literally type in or ask an AI assistant, not the questions that suit you. That difference determines whether you get cited or ignored. In B2B this means concretely:

Fit questions: “What type of company is this for?”, “Does this also work for a team of X people?” AI engines use this kind of content to match buyer and solution.
Comparative questions: “What is the difference between X and Y?” This is where much B2B purchase research begins and where AI likes to lift a clear answer.
Objection and risk questions: “What if it does not work?”, “How long does implementation take?” Honest answers here build trust, with the reader and with the model.
Price and process frames: you do not have to name an exact rate, but do give an honest frame so the answer is usable.

Phrase each question the way a human asks it, in full sentences, not as a keyword. And keep each answer short and complete: two to four sentences that stand on their own, without reference to “as above”. That self-containment is precisely what an AI engine needs to extract your passage. If you want to tune this well to your target audience, our guide on GEO for B2B will take you further.

What are the limits of FAQ schema?

FAQ schema is no guarantee of citations and certainly no substitute for good content. It is an amplifier, not an engine. Anyone who thinks a small block of JSON-LD suddenly triples visibility in ChatGPT is in for a disappointment. The figures on the direct effect contradict each other, and that says enough: there is no miracle cure.

What is always true: a page that genuinely contains the best answer to a question, written in clear self-contained blocks, wins. The schema makes that answer a bit easier to find and unambiguous. So never start with the schema. Start with your buyer’s question and with an answer that is worth citing. FAQ schema comes afterward, low effort, no risk, and does logical work within a broader GEO foundation. What that foundation looks like, you can read in our complete GEO guide, and how to specifically end up in Google’s AI Overviews.

The short summary

Google no longer shows FAQ rich results since May 2026, but that does not make FAQPage schema pointless. AI engines do not read your JSON-LD as data, but the tight Q&A structure the schema enforces is excellent to extract. For B2B it is cheap insurance: low effort, no risk, and it strengthens how AI understands your content. The real gain lies in the discipline underneath, not in the schema itself. Start with the questions your buyers really ask, give self-contained answers, and let the schema confirm them.

Want to know how your entire site is ready for AI search engines and which quick wins deliver the most for your leads? Check out our approach for visibility in AI search engines or plan your free intake.

Implementing FAQPage schema for AI citations (after Google's deprecation)