FAQPage Schema in 2026: What It Still Does — and What Counts Instead

FAQPage markup used to earn FAQ rich results in Google. That feature is gone. ChatGPT and Claude do not see the markup when they fetch a page. What the study points to instead is the visible text: questions as headings, followed by a direct answer.

What is FAQPage Schema?

FAQPage is a Schema.org type that describes a page containing a list of frequently asked questions and their answers. You add it as a JSON-LD script block that mirrors the visible FAQ content on the page.

Each question is represented as a Question entity with an acceptedAnswer. Whether a system reads this block at all depends on the system — see below.

Why FAQPage Schema Matters Less Than It Used To

Google

FAQ rich results no longer appear in Google Search since May 7, 2026. Since August 2023 they had already been limited to government and health websites. The main reason to add FAQPage markup for Google is therefore gone. Google’s general rules for structured data still apply to any markup you keep: “Don't mark up content that is not visible to readers of the page.”

ChatGPT and Claude

When ChatGPT and Claude fetch a page live, they remove the JSON-LD. In the Ramos study, a fact that existed only in the schema was never used in an answer. What these assistants work with is the visible text in the HTML. If an answer matters, it has to be on the page.

Gemini

Gemini (URL context via the API) receives the JSON-LD; the Gemini app did not open the URLs at all in the study. Where a model does see schema, it weighs it about as much as an ordinary sentence. The risk lies in contradictions: if the FAQ markup is outdated, Gemini may quote the outdated answer.

What is open

Perplexity, training crawlers such as GPTBot and ClaudeBot, and Google AI Overviews were not part of the study. We do not know whether they use FAQPage markup. The study also has limits: fictional companies, one fact per page, status September/October 2026.

What Counts Instead: Visible Questions with Direct Answers

In the study, pages phrased as questions were found and cited more often. The author attributes this to the wording in the text, not to the markup. In practice:

• Use the question as a heading. Put each question in an <h2> or <h3>, phrased the way people ask it.
• Answer directly below it. Start the first sentence with the answer, then add detail.
• Keep every fact visible. Prices, dates, conditions and names belong in the text, not only in the markup.
• Back up answers. In the GEO paper (Aggarwal et al., KDD 2024), citing sources, quotations and statistics increased visibility in AI answers by up to 40% in a test environment.

More on this in the guide Content Structure for AI.

If You Keep the Markup: JSON-LD Code Snippet

There is no need to remove existing FAQPage markup, but it has to match the visible text exactly. If you keep it, use this pattern. Place the <script> block inside your <head> or at the end of <body>.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How long does delivery take?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Orders placed before noon on a working day ship the same day. Delivery within Germany usually takes one to two working days."
      }
    },
    {
      "@type": "Question",
      "name": "Can I return an item?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Yes. You can return unused items within four weeks. Use the return label included in your parcel."
      }
    }
  ]
}
</script>

The questions and answers in this example are placeholders. Every name and text must appear word for word as a visible question and answer on the same page.

Where to Place FAQPage Schema

You can place the JSON-LD block in either location:

• Inside <head> — common and easy to maintain.
• At the end of <body> — also fine. Some CMS platforms (WordPress, Shopify) inject structured data here via plugins.

Do not place JSON-LD inside <noscript> tags. Google can process structured data that JavaScript adds to the page, as long as it is in the rendered HTML. For other systems this is not documented, so markup in the static HTML is the safer choice.

Step-by-Step (If You Keep the Markup)

1. Write the visible FAQ first. Each question as a heading, each answer directly below it. This is the part that ChatGPT and Claude read.
2. Mirror it in JSON-LD. Copy the snippet above and replace the example questions and answers with the exact text from the page.
3. Add the script to your page. In WordPress, use a plugin such as Yoast SEO or Rank Math, or a custom HTML block. In custom HTML, paste it into <head>.
4. Validate the syntax. Use validator.schema.org to check that the JSON is valid and the types are recognised.
5. Update both together. Whenever an answer on the page changes, change the markup in the same step — or remove the markup.

Common Mistakes

• Information only in the schema, not visible on the page. ChatGPT and Claude strip JSON-LD when fetching a page, so a fact that exists only there is lost to them. Google’s rules also say not to mark up content that readers cannot see.
• Markup differs from the visible text. Outdated FAQ markup is the typical case: the page was updated, the JSON-LD was not. In the study, Gemini Pro then gave the outdated answer 8 out of 20 times.
• Adding FAQPage just for rich results. Google no longer shows FAQ rich results. Adding the markup for that reason has no effect.
• Several FAQPage blocks on one page. Combine all questions of a page in a single FAQPage entity. Duplicate blocks are harder to keep in sync.
• HTML inside the answer text. Keep the text property plain text, identical to the visible answer. Avoid embedding <a>, <strong> or other tags in the JSON string.
• Invalid JSON. A missing comma or an unescaped quotation mark makes the whole block unreadable. Validate after every change.
• Forgetting to test after CMS updates. Plugins and theme updates can silently remove or alter structured data. Re-test after major updates.

Official Sources

• Rick Ramos (October 14, 2026) — Schema markup gives AI assistants nothing plain text doesn’t
• Google Search Central — Latest documentation updates (May 8 and June 15, 2026: FAQ rich results)
• Google Search Central Blog (August 2023) — Changes to HowTo and FAQ rich results
• Google Search Central — General structured data guidelines
• Google Search Central — Generate structured data with JavaScript
• Aggarwal et al. — GEO: Generative Engine Optimization (KDD 2024)
• Schema.org — FAQPage type definition
• Schema Markup Validator

Related Guides