On 7 May 2026, Google quietly switched off FAQ rich results in the search results page. Five days later, a client I work with asked whether we should rip FAQPage schema out of their site to free up engineering time. My answer surprised them. Keep it. Add more of it. The audience for that schema just changed, but the value did not. If anything, the value went up. This piece is the long version of that thirty-second answer, with the evidence, the mechanism, and the things you should actually do this week.
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. Service pages, product pages, a dedicated FAQ page, the bottom of blog posts. Most of those have FAQPage schema on them, added at some point 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 on the search page, and you got a small boost in click-through rate when it did.
| 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 |
Now Google has switched off that visible reward. The instinctive reaction is the one my client had. If Google no longer shows it, why bother maintaining it. And if engineering hours are tight (which they always are in a UK SME), pulling the schema out feels like a sensible cleanup.
It is not. 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 from the open web, and they 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
Precision matters here because the language got confused fast online. 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, and it did not say anything about how schema is used elsewhere in the search ecosystem. The official position is documented in Google’s structured data guidelines. Read it carefully and you will see that the schema spec at schema.org/FAQPage is unchanged, only the rich-result eligibility on the SERP has gone.
That distinction is the whole game. 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, and none of them have announced any plan to stop pulling from it.
Why FAQ schema didn’t lose value, just changed its main audience
If you think about what an AI Overview is actually doing when it answers a question, the appeal of well-structured FAQs becomes obvious. The system 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 that I have watched it happen across the BrisTechTonic client base for months.
If you want the broader theory of how AI search is reorganising what gets found, my piece on Generative Engine Optimisation covers the model end to end. FAQ schema is one slice of that picture. It is also the slice that just got cheaper to win, because most of the market is about to delete theirs.
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. They had not run a campaign. The only thing they had done in the previous quarter was tidy up their FAQPage schema on their main product pages, with our help. The schema was already there. We made it correct, kept the question wording natural, and made sure the answers were 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. The pages that ranked best in classic Google were not the same as the pages getting AI traffic. The pages getting AI traffic were the ones built around explicit questions. For the legal-services pattern, see our legal services industry page.
Both are anecdotes, not a controlled study. But they sit alongside something more uncomfortable. Across the same period, we have not seen a single AI tool deprioritise schema-marked Q and A. We have only seen the opposite signal: AI tools learning that this content is reliable and citing more of it. The platforms publishing about how their indexes work, including Cloudflare’s writing on AI bot management, keep treating structured content as a primary input. Nobody who matters in the AI search world has signalled that structured Q and A is going anywhere.
Should you remove your existing FAQ schema?
Almost certainly not. The two cases where I would consider removing it are narrow.
First: the FAQs themselves are poor quality, repetitive, or stuffed with keywords that misrepresent the page. That was a problem before the deprecation and it is a problem after. Fix the content, do not delete the schema.
Second: the schema is on pages where the visible FAQ content has been removed (sometimes during a site refresh, the schema gets left behind in the template). That creates a mismatch between schema and visible page, which has always been against the spec. Clean those up.
Everywhere else, leave it in place. The cost of keeping it is zero. The cost of removing it is the small, ongoing chance that the AI tools you actually want to be cited by quietly stop seeing the structure they prefer. If you want a sense of how often AI tools surface citations from specific structural patterns, my earlier piece on ChatGPT citing Reddit walks through the mechanism on a different source type. The principle is the same.
How to add FAQ schema that AI tools actually pick up
If you are starting fresh or rebuilding a section, there are five things I would do every time. None of them is technically hard. Most are about discipline rather than skill.
- Write the question the way your customers ask it. Not the way you wish they did. “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. Do not truncate to “see below” or “click for more.” The answer property in FAQPage schema 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. Mismatch is a long-standing schema sin and AI tools spot it as fast as Google ever did.
- Group five to ten FAQs per page, not fifty. Long FAQ lists dilute the signal. Pages with a focused cluster of high-intent questions earn more citations than pages that try to cover everything.
- Use canonical, clean JSON-LD. Do not nest it inside other structured data unnecessarily. Validate with Schema.org’s reference and Google’s Rich Results Test, even now. The test still runs even though the visible rich result is gone, and it is the fastest way to catch syntax errors that will quietly stop AI tools parsing your page.
That is it. Five steps, none of them expensive. Most of them are improvements to the underlying content as well as to the schema, which is why they pay off twice.
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 now. Sitelinks search boxes, video carousels, certain rich results in certain verticals. Each time, the temptation is to assume the underlying signal lost value because Google stopped rewarding it visibly. That is rarely true. The same data is still being parsed, scored, and used. It just stops being visible to the casual SERP-watcher.
For a wider read on what’s actually moving on the search results page in 2026, my BrightonSEO 2026 takeaways cover the structural shifts behind this kind of deprecation. The short version is that the modern SERP is increasingly a destination only for navigational and transactional queries, while informational queries (the kind FAQs answer) are moving wholesale into AI surfaces. The schema work you do today is positioning you for those AI surfaces, not for last year’s blue-link result.
Where to start this week
If you have ten minutes and want to do one useful thing, open one of your most important service pages, scroll to the FAQ section if it has one, and check three things. Does the schema match the visible content. Are the answers complete inside the schema, not truncated. Do the questions match the way your customers actually ask. If the answer to any of those is no, that is your edit.
If you have an hour, do the same on your top five revenue pages and note the gaps. If you have a day, audit your whole site for schema-page mismatches, missing canonical clean-up, and prose-only FAQ sections that have no schema at all. That last category is usually the biggest opportunity. Many older blog posts and service pages have natural Q and A structure that was never marked up because it predates anyone thinking about AI search.
If you would rather we did this for you, that is what our SEO packages include as part of an AI Search visibility audit. Either way, the brief is the same: 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. I run a small Bristol team helping UK SMEs stay visible as search moves into AI surfaces. See our SEO packages or book a discovery call.
