FAQ schema for AI search: does it help?

That will disappoint anyone who was sold FAQ schema as a shortcut to AI answers. Below, we do three things:

• Separate what Google documents from what vendors assert.
• Say what to do with markup you already have.
• Lay out a small test a service business can run.

What happened to Google’s FAQ rich result?

Google retired it. The Google Search Central changelog has a May 2026 entry. It says the FAQ rich result “will no longer appear in Google Search starting May 7, 2026.”

A 15 June 2026 entry says Google removed the feature’s documentation because it is no longer shown.

We read the changelog on 22 October 2026. Neither entry says what site owners should do with existing markup. The entries do not say whether FAQPage remains valid, and they do not mention Search Console reports.

The old rich result was the expandable question list under a blue link. Its removal ends a search appearance. It does not tell you how AI features pick sources, and Google’s documentation draws no line between the two.

Our FAQ schema glossary entry defines the term if you need it.

Does Google say FAQ schema helps AI Overviews and AI Mode?

No. Google’s AI features documentation (last updated 10 December 2025) says there are no additional requirements to appear in AI Overviews or AI Mode. You need no new machine-readable files or markup, and “there’s also no special schema.org structured data that you need to add.”

The one technical condition it names is ordinary: a page must be indexed and eligible to be shown in Google Search with a snippet. That makes indexing the first check on any page you hope to see cited. It is a cheaper fix than any markup project.

Google’s newer guide to optimizing for generative AI features (last updated 10 July 2026) goes further. It says structured data “isn’t required for generative AI search” but is still a good idea to keep using as part of overall SEO. It also lists practices you can ignore:

• Creating llms.txt or other special machine-readable files.
• Chunking content into tiny pieces for AI.
• Rewriting content in a particular style just for generative AI.
• Overfocusing on structured data.

“Overfocusing on structured data” sits right beside llms.txt on that list. We read that as Google saying markup is fine to keep and is not where AI visibility comes from.

What do Bing and other AI systems say?

Bing documents visible content, not schema. In its AI Performance announcement (10 February 2026), Microsoft says clear headings, tables and FAQ sections help surface key information. The post does not mention schema markup.

That sentence treats FAQ sections as content a person and a crawler can both read. It doesn’t claim FAQPage structured data raises citations.

What does the Bing report measure?

The same post introduces a Bing Webmaster Tools report. It counts citations in AI-generated answers, which gives you a way to measure.

What about other AI assistants?

For other assistants, we found nothing. We did not locate primary documentation from OpenAI, Perplexity or Anthropic.

None ties FAQPage markup to how they choose sources, so we treat any such claim as unproven. Our post on how AI assistants choose sources covers what is documented.

• GoogleNo special schema.org markup is needed for AI Overviews or AI Mode.
• BingDocuments visible content, such as clear headings, tables and FAQ sections. No mention of schema.
• Other assistantsWe found no primary documentation tying FAQPage markup to how they choose sources.

The dates below show how the documented picture built up.

1. 10 December 2025Google's AI features documentation, last updated. No special schema is needed.
2. 10 February 2026Microsoft's Bing AI Performance announcement names FAQ sections as helpful content.
3. 7 May 2026The Google FAQ rich result stops appearing in Search.
4. 15 June 2026Google removes the feature's documentation.
5. 10 July 2026Google's guide to optimizing for generative AI features, last updated.

Which FAQ schema claims are documented, experimental or theory?

We label AI-search claims in three tiers:

• Documented: a vendor says it in official documentation.
• Experimental: someone has tested it, and results are partial.
• Theory: plausible, untested.

The table sorts the common FAQ schema claims accordingly, as of 22 October 2026.

On that last row, Google’s featured snippets documentation was last updated 10 December 2025. It says Google systems choose snippets automatically. It also says site owners cannot mark a page as one.

We opened no equivalent page for question boxes, so we make no claim about them. Our People Also Ask post covers what is known.

Is the visible FAQ still worth writing?

Yes, when the questions are real. A visible FAQ answers objections before a sales call and hands crawlers short, self-contained passages. Neither benefit needs markup.

What changed is the reason to write one: forget the search decoration and answer what a buyer actually asks.

Here is how we decide whether a page gets an FAQ block.

1. Keep it if you can list at least four questions that came from calls, forms, chat logs or Search Console queries.
2. Skip it if the page above already answers every question. Repeating yourself adds length, not depth.
3. Put the direct answer in the first sentence, then the caveat. Answers of 40 to 90 words are easy to lift without losing meaning.
4. Make each answer stand alone. It should name the service, the place and the condition, because an extracted answer travels without its heading.
5. Drop questions that only flatter you. “Who is the best plumber in Phoenix?” invites an answer nobody can honestly give about themselves.

Our posts on writing FAQs for answer engines and on FAQ pages and SEO go through the format in more detail.

What does an answer block look like before and after?

A usable answer block states the answer, the condition and the next step in the first two sentences. Below is a hypothetical example for an invented plumbing company. It is illustrative, not a measured result, and it was not tested in any AI Overview.

Before (generic):

Our experienced team is dedicated to providing high-quality service and customer satisfaction. Contact us today to learn more about our estimates and how we can help with your plumbing needs.

After (answer first):

We give a written estimate for repairs before work starts, and you approve it first. If the problem needs a diagnostic visit, the fee is quoted on the phone and applied to the job when you go ahead. Larger replacements get an on-site quote, usually within the same visit.

The first version could describe any contractor. The second names a policy, a sequence and a limit, so an assistant that lifts only that paragraph still says something true about this company. That property is worth chasing with or without markup.

Here is how to turn a generic answer into an answer-first one:

1. State the answer in the first sentence.
2. Add the condition or limit in the second sentence.
3. Name the next step the reader should take.
4. Name the service, the place and the condition, so the answer still makes sense without its heading.

We do not have a dated before-and-after capture of these blocks rendering in AI Overviews or in question boxes. Capturing one is the first step of the test below, and until it exists the claim stays at “plausible”.

What should you do with FAQ markup you already have?

Leave accurate markup alone, and delete markup that is templated, hidden or stale. Google’s structured data policies say structured data must be a true representation of the page. They also say not to mark up content readers cannot see.

A structured data issue can lead to a manual action, which removes rich result eligibility. With the FAQ rich result retired, that risk is mostly a hygiene issue now. We check, in order:

1. Is the markup in the visible text? If a question or answer exists only in JSON-LD, delete the markup or put the text on the page.
2. Does the visible answer match? After any edit to the FAQ copy, the schema must change with it.
3. Is it the same list on many pages? If the same ten questions appear on fifty pages, cut to the questions each page truly answers.
4. Is it hand-maintained or generated? A plugin that regenerates markup from page content is fine. A pasted block nobody updates will drift.
5. Would removing it cost money? If it needs developer time, bundle it with other work. Nothing here is urgent.

Here are the same five checks as a table.

For an overview of which structured data types still matter to Google, see our schema markup guide and the post on whether schema helps SEO.

Did FAQ schema help one of our own projects?

One of our own projects used FAQ markup at scale. The B2B case study records FAQ schema on 343 posts. It also records 4,802 Google clicks in September 2026 against 1,637 in April 2026 (Google Search Console).

Other work happened in that period, so it isn’t a test of FAQ schema, and we don’t credit the markup for the change.

How can a service business test FAQ schema for AI answers?

Run a controlled test with a baseline, a control group and a fixed review date. AI answers vary by query, location and session, so one screenshot proves nothing.

The design below is our method, offered as a hypothetical plan. We have not run it on a client site yet.

1. Pick the pages. Choose ten service or location pages with similar traffic and the same template. Confirm each is indexed in the Page indexing report in Search Console, since Google says a page must be indexed to be eligible.
2. Write five target questions per page. Take them from your calls and Search Console queries, and log them in a sheet with the date and exact wording.
3. Record a baseline. For each question, search in a private window on three different days.
   • Note whether an AI Overview appears, whether your domain is cited, and who is.
   • Google’s changelog (August 2024 entry) says AI Overviews are counted in the Search Console Performance report, so log impressions and clicks too.
   • For Bing, use the AI Performance report in Bing Webmaster Tools.
4. Split the pages into three groups. Changing one thing per group is what lets you tell them apart.
   • Three pages stay untouched as the control.
   • Four get a visible answer-first FAQ block with no markup.
   • Three get the same block plus FAQPage markup.
5. Wait, then repeat. Hold for at least six to eight weeks, which is our rule of thumb rather than a documented figure. Repeat the manual checks on three days and compare against the baseline sheet.
6. Decide with a threshold you set first. For example: “keep the block if citations or clicks rise on the test pages and not on the controls.” Write the rule before you see the numbers.

1. Pick ten pagesChoose similar service or location pages that are all indexed.
2. Write five questions eachTake them from calls and Search Console queries.
3. Record a baselineSearch each question on three days and log what appears.
4. Split into three groupsControl, answer block only, answer block plus markup.
5. Wait, then repeatHold six to eight weeks and run the same checks.
6. Decide by a rule set firstWrite the threshold before you see the numbers.

Do the arithmetic: ten pages with five questions, checked on three days, make 150 observations per round. AI answers shift between searches. A positive result is a hint.

A null result is likelier, and it still tells you not to spend a quarter on markup.

Our GEO measurement limits post explains why AI visibility numbers need this much care.

What does this mean for your service business?

The documented payoff for FAQ markup ended on 7 May 2026. For you, that means nothing here is urgent. Leave accurate markup alone.

Put your time into the visible answers instead. Start with the real questions your customers ask, and put the direct answer in the first sentence.

If a vendor pitches an AI visibility project built on schema, ask for the test. The plan above gives you a template: ten pages, five questions each, three groups.

What does the data not show?

It does not show that FAQ schema helps AI answers, and it does not show that it hurts them. The documented picture is narrow and fully on Google’s side: one rich result retired, and one statement that no special markup is needed.

The specific gaps:

• No primary source we opened links FAQPage markup to AI citation on any platform.
• We have no dated capture of before-and-after answer blocks in AI Overviews or question boxes, and we do not claim one.
• Google’s documentation does not say what happens to existing FAQPage markup after retirement.
• We did not retrieve schema.org’s own FAQPage definition during this check, so we make no claim about its status.
• Our B2B client figures were not produced by an FAQ experiment.

Anyone selling a number for “AI citation lift from schema” owes you a method, a sample and a date. Our post on LLM SEO myths lists the common unsupported ones.

Where should the effort go instead?

Put it into the things Google names in its documentation:

• Make sure key pages are indexed and snippet-eligible.
• Write content with a point of view that is not available elsewhere.

Google’s AI guide calls this non-commodity content. It means unique expert or experienced takes that go beyond common knowledge.

For a service business, that usually means pages that state your pricing logic, your limits, service areas and what happens after the call.

Our guide to ranking in AI Overviews walks through the documented factors. Our AI search optimization service covers the work we do on pages, entities and measurement.

And if someone proposes a large AI visibility project built on schema, ask for the test.

Frequently asked questions

Check how your key pages look to a crawler with our website SEO checker, then Get Your Organic Growth Plan.

Sources

• Google Search Central documentation updates (changelog), Google, FAQ rich result entries May and June 2026, accessed 22 Oct 2026
• AI features and your website, Google Search Central, last updated 10 Dec 2025, accessed 22 Oct 2026
• Optimizing for generative AI features on Google Search, Google Search Central, last updated 10 Jul 2026, accessed 22 Oct 2026
• General structured data guidelines and policies, Google Search Central, accessed 22 Oct 2026
• Featured snippets and your website, Google Search Central, last updated 10 Dec 2025, accessed 22 Oct 2026
• Introducing AI Performance in Bing Webmaster Tools, Public Preview, Bing Webmaster Blog, 10 Feb 2026, accessed 22 Oct 2026