FAQ schema after Google removed FAQ rich results: what it still does

What is FAQ schema?

FAQ schema is a block of structured data that uses the FAQPage type from schema.org. It lists the questions and answers on a page in a form machines can read directly. Most often it is written as JSON-LD, a script in the page’s code that visitors never see.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How long does a crown take?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Two visits, about two weeks apart."
      }
    },
    {
      "@type": "Question",
      "name": "Do you take weekend appointments?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Yes, Saturdays from 9 to 1."
      }
    }
  ]
}
</script>

For years, the reason to add this was visual. Google could show the questions as expandable rows under your search result, taking up more of the page. That display is gone, so that reason no longer holds.

What did Google change, and when?

The change came in two steps, nearly three years apart. Google’s own documentation changelog and blog record them.

So FAQ schema no longer earns extra space in Google, whatever older screenshots show: the feature itself is gone.

HowTo went first. Google’s 2023 post cut those results back too, and they later stopped appearing altogether. If you still have HowTo markup, it is in the same position.

Google’s Q&A type looks like a replacement, but it is not one. It is meant for pages with one question followed by its answers, such as a forum thread where people submit replies. Do not relabel a business FAQ as Q&A in the hope of getting the old display back. Our guide to schema markup lists the types that still produce something in Google.

What does FAQ schema still do?

Less than it did, but it is not worthless. Here is what each major company actually says.

Google Search. No rich result. In its 2023 post Google said of such markup that “there’s no need to proactively remove it” and that structured data not being used “does not cause problems for Search, but also has no visible effects”. Its guide on AI features, updated July 2026, says that structured data is not required for generative AI search and that there is no special schema to add.

Microsoft. In an October 2025 post from its Bing team, Microsoft says schema “can label your content as a product, review, FAQ, or event”, turning plain text into data machines can interpret. It also recommends question and answer pairs, saying assistants can often lift them word for word. This is the clearest support any major company gives for the markup today.

OpenAI, Anthropic and Perplexity. Their crawler and search pages do not mention schema at all, as read in November 2026. That neither confirms nor rules out any effect. A bot reading your HTML can see the block, and no one has published whether it counts.

So the honest position is that FAQ schema is a cheap, accurate label. It costs little, it does no harm, it may help some readers, and nobody can show it earns citations. Any claim stronger than that goes beyond what these companies have said.

Should you keep, add or remove it?

Most of the time the answer depends on how much work it takes.

If a plan you are paying for includes FAQ schema so that your pages stand out in Google, it is worth asking what that item is meant to do now.

Why do the visible answers matter more than the markup?

Because the answers are the part every system can use. A model that reads your page gets the words whether or not they have a label, and the same passage works for a person.

Three sources agree on this. Google says structured data must match the visible text. Microsoft says assistants can often lift question-and-answer pairs word for word. And the GEO research paper, posted in 2023, found that the content of a passage, such as added statistics and quotations, changes how often it is used in AI answers.

The sample we scanned shows where the gap really is. When we tested the MWY checker on 36 business websites in seven countries on November 3, 2026 (a hand-picked sample, not a market figure), it could read 32. Of those, 12 had questions and answers with FAQPage markup, 15 had questions and answers without it, and 5 had no such section we could detect. So most had already done the writing, and the markup was the smaller gap. The check is a signal: it can see that questions exist, not whether the answers are any good.

To write answers worth reading:

• Use real questions. Take them from your inbox, phone calls and reviews, in your customers’ words.
• Answer in the first sentence. “Two visits, about two weeks apart” and then the detail.
• Use real numbers. Prices, timeframes, limits, areas served.
• Keep each one self-contained. Do not write “as mentioned before”. The answer may be read alone.
• Put them where the question comes up. A question about crown prices belongs on the crowns page more than on a general FAQ page.
• Skip the marketing questions. “Why are you the best?” helps nobody, and readers can tell.

Our guide to answer engine optimization goes deeper into how to find and write these answers.

Which pages should carry questions and answers?

A single FAQ page is the common habit, but the answers often do more work next to the decision they belong to.

What does a good answer look like?

The difference is usually a number or a limit that the weak version leaves out. The examples here are invented for illustration.

The stronger versions can be quoted by a machine and used by a person, and they cannot be copied by a competitor without being true of that competitor too.

How many questions does a page need?

No company publishes a number, and there is no reason to hit one. As a rule of thumb, our checker’s advice is 8 to 10 questions across the site at first, each answered in 2 to 4 sentences. Stop when you run out of questions people really ask. Five real ones beat fifteen invented ones, and a question nobody has ever asked you is usually a sales line in disguise.

How do you audit the markup you already have?

This takes a few minutes per page:

• Search the page source for “FAQPage”. If it appears twice, a theme and a plugin are probably both adding it.
• Compare the markup with the page. Every question and answer in the block should appear on the page in the same words.
• Check that the answers are still true. Prices and hours written in a block are easy to forget.
• Look for the same block on many pages. If every service page carries the same questions, keep them on one page and write specific ones for the rest.
• Find out which tool writes the block. When a plugin stores answers apart from the page, someone can edit the page while the markup keeps the old wording. Note the tool, so the next editor changes both.

Do accordions hide your answers from AI?

Often a FAQ is built as an accordion, with answers that open on a click. Microsoft warns against this directly: “Don’t hide important answers in tabs or expandable menus”, because AI systems may not render hidden content.

What decides the risk is how the accordion is built. Check it yourself: open the page, choose View Source (not Inspect), and search for the text of one answer.

• The answer is in the source. It is only hidden by styling. A bot reading raw HTML has the text, although Microsoft’s warning suggests that hidden text may not be treated the same.
• The answer is not in the source. A script loads it on click. A bot that does not run JavaScript never sees it, while your markup may still state it, leaving a mismatch.

For your most important answers, the safe choice is to leave them open, with no click at all. A long page of open answers also reads fine on a phone if the questions are clear headings.

How do you add FAQ schema?

• WordPress. The FAQ blocks in Rank Math and Yoast SEO add the markup for you when you write your questions in them.
• Other platforms. Check whether your theme offers a FAQ section with markup, and view the source to see what it outputs. If your platform lets you add code to a page’s head, paste the JSON-LD there.
• By hand or from a developer. Use the block at the top, and have the code generated from the same text the page shows, so the two cannot drift apart.

Then check the code at validator.schema.org. Google’s Rich Results Test is no longer the right tool for this type, because there is no rich result left to test for.

Common mistakes to avoid:

• Answers in the markup that are not on the page. Google’s policy is not to mark up content readers cannot see, and a violation can lead to a manual action.
• The same FAQ on every page. It adds nothing and reads as repetition. Keep each block specific to its page.
• Sales copy instead of answers. The markup does not change what the words are worth.
• Answers that go stale. Prices and hours in a FAQ age quickly, so review them along with the rest of the page. A good moment is whenever you change a price or your hours anyway.

How the MWY checker scores FAQ schema

The MWY checker has one item for this among its 24, with a weight of 6. Its rule: a questions and answers section with FAQPage markup gets a pass, the section without markup gets a partial, and no section gets a fail. Our methodology page lists every weight.

A partial earns half the weight, so the section and the markup each carry half of this item. We kept the markup in the pass condition because it states questions and answers explicitly for any machine that reads it and Microsoft names it, and we say here plainly that this is a judgment on our part: it has nothing to do with Google’s old display, and we have no evidence that it earns a citation. The item is marked as a signal, because a checker can see that questions exist, not whether the answers are useful. A pass does not promise that an assistant will quote you.