Google deprecated FAQ rich results in May 2026. Five days later, a client asked if we should rip out their FAQPage schema. The answer is no, and the reason matters.
Why this matters if you have FAQs on your site
If you run a UK business with a website older than two years, you almost certainly have FAQ sections somewhere, with FAQPage schema added by an agency or a plugin like Rank Math or Yoast. For years the deal was straightforward: you marked up your questions and answers, Google sometimes showed them as a rich result, and you got a small boost in click-through rate.
| Schema type | Status in Google rich results (2026) | Still useful in 2026? | Why |
|---|---|---|---|
| FAQPage | Deprecated as SERP rich result for most sites | Yes, extremely | Now read by LLMs for extractable Q and A blocks |
| HowTo | Deprecated as SERP rich result | Yes for AI extraction | Numbered steps get lifted into AI answers |
| Article | Live and used for Top Stories | Yes | Feeds dateModified freshness signals to AI |
| Organization / LocalBusiness | Live and used for knowledge panel | Yes, foundational | Anchors entity trust for AI citation |
Pulling that schema out today is one of the easier ways to make yourself invisible to ChatGPT, Perplexity, Claude, and Google's own AI Overviews. Those tools have spent two years training on structured Q and A content and pull from it constantly when answering live queries. The deprecation closed one tap. It did not close the other, much bigger one.
What Google actually deprecated on 7 May 2026
Google deprecated the visible rich result, the expanding accordion of questions that used to appear under some search listings. It did not deprecate FAQPage schema itself. The schema spec at schema.org/FAQPage is unchanged, only the rich-result eligibility on the SERP has gone. The schema is the underlying machine-readable structure. The rich result was one consumer of that structure. AI search tools are a different, larger, more aggressive consumer of the same structure.
Why FAQ schema didn't lose value, just changed its main audience
When an AI Overview answers a question, it has to find, parse, score, and combine multiple sources to write a single answer paragraph. It strongly prefers sources where the question is explicit, the answer is contained, and the relationship between the two is unambiguous. A page with proper FAQPage schema gives the AI a clean, labelled, machine-readable pair. A page with the same content as a wall of prose gives the AI something it has to infer. In a market where dozens of pages might address the same query, the marked-up version wins the citation. Not always. But often enough.
What we're seeing across BrisTechTonic client sites
One of our precision-engineering clients started appearing in multiple AI Overviews this month for high-intent searches in their sector. They had not changed their site or run a campaign. The only thing they had done in the previous quarter was tidy up their FAQPage schema on their main product pages, keeping the question wording natural and the answers complete inside the schema rather than truncated.
A separate client, in legal services, has been quietly accumulating AI referral traffic that does not show up in the usual SEO dashboards because the visit lands without a search-engine referrer. When we cross-checked their highest-cited pages, the common factor was structured Q and A content with FAQPage schema. Both are anecdotes, not a controlled study, but across the same period we have not seen a single AI tool deprioritise schema-marked Q and A.
Should you remove your existing FAQ schema?
Almost certainly not. Two narrow cases where I would consider removing it: first, the FAQs themselves are poor quality, repetitive, or keyword-stuffed (fix the content, do not delete the schema); second, the schema is on pages where the visible FAQ content has been removed, creating a mismatch. Everywhere else, leave it in place. The cost of keeping it is zero.
How to add FAQ schema that AI tools actually pick up
- Write the question the way your customers ask it. Do you do PPC for small businesses? beats Pay-per-click services scope. AI tools match on natural phrasing.
- Answer the question inside the schema, completely. The answer property should be a self-contained reply, ideally one to four sentences, that stands on its own.
- Match the schema to the visible page. Whatever is in the schema must also be on the page in human-readable form, in the same place, with the same wording.
- Group five to ten FAQs per page, not fifty. Long FAQ lists dilute the signal.
- Use canonical, clean JSON-LD. Validate with Google's Rich Results Test, which still runs even though the visible rich result is gone.
The bigger picture: signals do not die, they migrate
The 7 May change is not unusual. Google has been steadily removing visible search-page features for two years. Each time, the temptation is to assume the underlying signal lost value. That is rarely true. The same data is still being parsed, scored, and used. The modern SERP is increasingly a destination only for navigational and transactional queries, while informational queries are moving wholesale into AI surfaces. The schema work you do today is positioning you for those AI surfaces.
Where to start this week
Open one of your most important service pages, scroll to the FAQ section, and check three things: does the schema match the visible content, are the answers complete inside the schema, and do the questions match the way your customers actually ask. If the answer to any is no, that is your edit. Do not rip out your FAQ schema because Google stopped showing it. Tidy it, extend it, and let the next-generation search tools cite you for the next two years.
By Chris McDowell, founder of BrisTechTonic, and Olivia, Technical SEO lead at BrisTechTonic.