FAQ Schema Did Not Die on 7 May 2026 — The Rich Result Did. Here’s What That Means for Your Visibility.

On 7 May 2026, Google removed FAQ rich results from search. The FAQPage structured data type itself is not deprecated. Those are different things, and the difference determines what you should do next. The Search Console FAQ rich result report and the Rich Results Test FAQ check are scheduled for removal in June 2026; the Search Console API support follows in August 2026. Sources: Google Search Central developer documentation, 7 May 2026; Barry Schwartz, Search Engine Land, 8 May 2026.

The reaction has been larger than the change. Within twenty-four hours my LinkedIn and X feeds filled with declarations that schema is dead, that FAQs were always a mistake, and that anyone who told you to add FAQ schema for AI visibility owes you an apology. None of that is true, and most of it is unhelpful — particularly to the people who actually need to make a decision this week about what stays and what goes.

This is a practitioner read. What changed, what did not, why it happened, who it affects, and what to do — segmented by the kind of business you actually run.

What changed on 7 May 2026 — and what did not

Two facts, separately.

What changed. FAQ rich results — the expandable question-and-answer dropdowns that used to appear beneath certain search listings — are no longer shown in Google Search. They had already been restricted to authoritative government and health sites in August 2023, so for most websites this was the formal end of a feature that had been functionally absent for almost three years. Search Console reporting and Rich Results Test support follow in June 2026; the Search Console API support is removed in August 2026.

What did not change. FAQPage structured data is still valid markup. Google’s own documentation, updated 7 May 2026, states that you do not need to remove it — structured data that is not being used does not cause problems for Search. At the crawler layer, the schema is parsed as machine-readable structured content by Bingbot (whose index is one of the retrieval sources used by ChatGPT search and Microsoft Copilot grounding), PerplexityBot, voice-assistant indexers, and the retrieval-augmented generation crawlers that index the open web. What each of those systems then does with the parsed data when generating an answer — whether it stores it, retains it, or weights FAQPage as a feature rather than treating it as plain content — is not publicly documented for any major AI system except Google’s own.

If you only read one paragraph, read this one: the SERP feature is gone, the schema type is not. They are not the same thing, and conflating them is what the next two months of bad advice will be built on.

Source: Google Search Central FAQPage documentation, last updated 7 May 2026.

The history that explains why this was always coming

FAQ rich results launched in May 2019. For about two years they were the cleanest free SEO win available — add a few lines of JSON-LD, and your single listing could occupy three or four times the SERP real estate of a competitor’s. Practitioners reasonably saw the opportunity and used it. Then the predictable thing happened.

In August 2023, Google restricted FAQ rich results to well-known authoritative government and health websites. Across tracking tools, FAQ rich result impressions for non-authority sites collapsed within weeks. The feature did not disappear — eligibility did. For most of the SEO industry, the rich result has been functionally dead since late 2023.

What changed on 7 May 2026 was administrative. The eligibility window stayed open for the eligible sites, but the feature itself stopped firing. Google’s March 2026 core update extended the cut: across tracked sites, FAQ rich result impressions fell again from the post-2023 baseline. By April 2026 the feature was generating almost no impressions for almost anyone. The May 2026 announcement closed the door on a feature that was already mostly closed.

The two-step sequence — restrict eligibility, then remove the feature entirely — is not new. Google did the same thing with HowTo rich results, deprecated entirely on desktop in September 2023 after a period of mobile-only restriction. The pattern reads as a Google policy: when a structured data feature gets aggressively scaled by SEO tooling and stops faithfully describing the page, the rich result is withdrawn before the markup is. The schema layer remains as something machines can still parse if it accurately describes what the page actually contains.

Lily Ray (VP, SEO & AI Search at Amsive; founder of Algorythmic) noted on LinkedIn on 8 May 2026 that this fits a longer pattern she had warned about in a 2019 Moz article: when too many sites use the same exact technique for SEO/GEO, Google steps in. The 2019 piece is worth re-reading in this context because it was both the article that helped popularise FAQ schema adoption and the article that included an explicit “Risks involved with implementing Schema” section warning that misuse at scale could cost the industry the rich result. The warning was issued from inside the same piece that taught people how to add the markup. Whether you read the May 2026 outcome as cause-and-effect or as Google solving for AI competition is a question we will come back to. The behavioural pattern is the same either way: a markup type that becomes a scaled extraction shortcut tends to lose its rich result before it loses its validity.

Where FAQ schema actually sits — the four-floor view

The mistake driving most of this week’s commentary is treating FAQ schema as a single thing. It is not. It has always been two layers conflated: a SERP feature (the rich result) and a content description (the structured data). One layer was removed. The other was not. The reason this matters is that the two layers do different jobs, on different floors of the visibility model.

Read in order: Floor 1 is your entity foundation — consistent business identity across Wikidata, Companies House, Bing Places, Google Business Profile, and your Organization schema. Floor 2 is content extractability — whether your pages can be parsed cleanly by both Googlebot and AI retrieval systems, with definitions, named statistics, attributable claims, and proper heading hierarchy. Floor 3 is third-party trust and selection — editorial citations, named-source corroboration, the signals that determine whether you are eligible to be recommended at all. Floor 4 is the agentic layer — the layer being built now, where AI agents transact on a buyer’s behalf via emerging standards like WebMCP.

FAQ schema lives on Floor 2. It is one tool among several for making content extractable. The rich result that used to appear in Google’s SERP was a third-party convenience — Google reading your Floor 2 markup and converting it into a Floor-3-adjacent visual feature. That bridge has been removed. Floor 2 itself has not.

The temporary ladder visual in the building above is doing real work. For a window of about six years, FAQ schema was the easy ladder between solid Floor 2 content and Floor 3 visibility. That ladder is now gone. The lift — AI Overviews, AI Mode, Perplexity, ChatGPT search, Gemini grounding — is being built. The floors remain. The way you reach the higher floors has changed; the building has not moved.

Why now? Five theories worth taking seriously

Anyone telling you they know exactly why Google made this change on this date is overclaiming. There are several plausible reasons, and the practical answer for what to do is the same regardless of which is correct.

Theory 1 — spam saturation forced eligibility removal. Lily Ray’s argument: an industry-wide push for FAQ schema is critical for GEO produced an influx of templated FAQ schema across 168,000+ articles (her count). In Ray’s framing, anything that can be spammed in SEO will be spammed, and Google’s response is the same as the 2019 launch trajectory: open feature, observe spam, withdraw feature. This is the most internally consistent theory and matches the 2023 HowTo precedent.

A concrete example of how this pattern works in real time: across LinkedIn and SEO industry blogs in 2026, a specific claim has been circulating that pages appearing in Google AI Overviews are 3.2x more likely to have FAQ schema implemented. The citation given for that claim is a Search Engine Land article URL that returns a 404. No primary source from OpenAI, Anthropic, Google, or Perplexity confirms any specific multiplier for FAQ schema in AI Overview citation. The absence of a verifiable source has not slowed the claim’s adoption. This is the spam pattern starting again under a new acronym — an unverified statistic becomes industry consensus, mass implementation follows, and the reward lever gets removed when the pattern is no longer useful to the system that was rewarding it.

Theory 2 — AI retrieval no longer needs the schema layer. Bree Sharp (local SEO strategist) made a sharper version of a point that has been quietly true for some time: large language models are trained on natural language and parse structured data far less rigidly than Googlebot does. A clean Q&A in plain prose — under a clear question heading, with a self-contained answer — does the same job for AI extraction without the schema overhead. If Google’s evaluation pipeline now reads pages the same way, the schema layer is no longer giving Google information it could not get anyway.

Theory 3 — AI Mode push. Pranav Labdhe’s read on X: removing FAQ rich results pushes users toward Google AI Mode, where suggested follow-up questions inside AI Mode (rather than expandable dropdowns in classic SERPs) become the conversational pattern. This is plausible because AI Mode interest spiked in Q1 2026 and Google has been rebalancing classic SERP features (HowTo, FAQ, news carousels) toward AI Overview consolidation since the September 2024 AI Overview rollout.

Theory 4 — competitive asymmetry. William Wright’s argument on X: Google can confidently extract Q&A content from prose, while smaller AI competitors cannot — or are slower to. Removing the schema feature widens Google’s relative advantage in AI summarisation. Plausible but speculative; the timing is suggestive but no internal Google document supports it.

Theory 5 — quality-update alignment. Several practitioners noted that aggressive FAQ scaling correlated with sites impacted by Google’s Helpful Content Update sequence. If FAQ schema spam is an HCU pattern indicator, removing the rich result is consistent with Google’s broader push toward content that demonstrates substance rather than gaming markup hooks.

Pick whichever combination feels true to you. The action that follows is the same in every case: stop using FAQ schema as a trick and start treating it as what it always should have been — an honest description of pages that genuinely contain question-and-answer content.

Who is actually affected, and how

The practical impact is not uniform. The framing FAQ schema deprecation hits everyone is exactly the kind of generic advice that wastes a marketing director’s afternoon. Here is the segmented read.

Two specific cases worth naming, because they show how different the same announcement looks from different angles.

A regulated UK law firm with FAQ pages on a specific area of practice — say, drug driving or matrimonial finance — is barely affected. The FAQ rich result was already withdrawn in 2023 for non-government, non-health sites; the page was already ranking on the strength of the content, internal linking, and entity signals. Removing the schema would be unnecessary work for no benefit: the JSON-LD accurately describes what is on the page, and at the crawler layer it remains a clean structural signal about which text is question and which is answer. Google’s own guidance is that you do not need to remove it. Schema stays. Templated boilerplate FAQ blocks bolted onto unrelated service pages should come off.

A B2B SaaS platform with a hundred templated FAQ blocks across its marketing site — the same six questions about security, compliance, pricing, onboarding rephrased — is in the opposite position. The rich result was a notional benefit; the templated structure is a quality liability. Strip the FAQ schema from pages where the FAQ block is filler. Keep it on the actual help-centre pages and on the FAQs that capture real, specific buyer questions. Move the rest of the effort into CITATE-aligned evidence layers — named statistics, attributable claims, source-cited paragraphs.

The mistake to avoid is the global one: stripping all FAQ schema across the estate to clean up. That is treating a content-quality decision as a markup-removal task, and it deletes the AI extraction signal alongside the dead SERP feature. Schema follows content. Decide about the content first.

Should you keep your FAQ schema? A three-question test

The decision is local, page by page. Three questions answer it.

Question 1 — Is the FAQ visible on the page? If the questions and answers do not appear in the rendered HTML for human visitors, the schema is decorative. It was added for the rich result. The rich result is gone. Remove it. Google’s own guidelines have always required that the FAQ content be present and visible on the page. Pages that quietly violated that rule were never robustly eligible anyway.

Question 2 — Are the Q&As genuinely useful? Real questions real users ask, with answers that genuinely answer them, ideally under fifty words each, with no marketing filler. If the questions read like keyword research output and the answers read like product copy, rewrite them. The schema follows the content; if the content is bad, fixing the schema does nothing.

Question 3 — Is this FAQ block templated across many pages with similar wording? If yes, audit. Keep on the pages where the Q&A genuinely fits the page’s subject. Remove or consolidate on the pages where it is filler. Templated FAQ blocks across a service-page footer were always the weakest pattern; this is the moment to clean them up.

If a page passes all three — visible, useful, not templated — keep the schema. It is an honest description of the page, and AI systems, Bing, voice assistants, and RAG pipelines continue to consume it. The cost of keeping it is zero. The cost of removing it is one extraction signal lost across multiple AI consumers, for no offsetting benefit.

There is a separate version of this question for sites built around FAQ pages as the actual format — help centres, regulator FAQ pages, government information sites. For those, the FAQ schema is not a layer added on top of content. It is the content. Keep it. The deprecation does not affect them in any structural way; only the visual SERP feature changed.

Why the abuse pattern was structural — and what honest FAQ markup looks like

Step back from the SEO industry behaviour for a moment and the deeper question becomes clear: the abuse pattern was partly structural. FAQPage was designed for pages whose primary purpose is question-and-answer content. The much more common case — a page about a product, service, article, or piece of legislation with an FAQ section attached — had no honest markup alternative. Publishers chose misrepresentation because the rich result reward was on that side of the choice. Joost de Valk diagnosed this clearly on 11 May 2026 and filed a FAQSection proposal at schema.org (issue #4816) the day after, with a companion proposal for a clean answer property to replace acceptedAnswer. The full structural treatment of why this matters beyond FAQ markup — including the Schema Half-Life Pattern, how schema vocabularies actually evolve, and the broader shift from syntactic validation to semantic credibility — is in the keystone piece: Schema Architecture for the AI Era.

What to do — this week, this month, this quarter

If you only have an hour this week, do this.

This week (1–2 hours). Export your historical FAQ rich result performance data from Google Search Console before the June 2026 reporting removal. Run a 28-day before/after comparison on pages that previously showed FAQ dropdowns — CTR delta is the only direct measurable impact of the 7 May change, and it is mostly only relevant for the regulated enterprises that actually still had the feature firing. Identify the pages on your site that have FAQPage schema. Most teams do not have an accurate inventory; this is a sitemap-wide audit using a crawler that parses JSON-LD.

This month (a focused half-day). Run the three-question test on every page with FAQ schema. Categorise into Keep / Audit / Rewrite / Remove. Implement the Remove decisions immediately — these are the pages where the schema was decorative. The Audit and Rewrite categories become content-improvement tickets, not schema-removal tickets. If you removed FAQ schema from templated service-page footers, replace it with nothing. The page does not need a Q&A block to rank.

This quarter (the bigger reframe). Move the effort that previously went into FAQ-schema-for-CTR up the floor model. Floor 1: confirm Bing Places listing, Wikidata entry where applicable, Google Business Profile consistency, NAP alignment across structured directories. Floor 2: implement CITATE standards on the pages that matter — declarative opening, defined terms inline, named statistics with sources, attributable claims, named entities, structured data that honestly describes the page. Floor 3: identify the twenty editorial outlets that cover your sector and start the long, slow programme of becoming a named source in their coverage. The University of Toronto’s September 2025 study across thirteen industries found that 92.1% of Google AI Overview citations come from earned media, not owned content; the Muck Rack Generative Pulse July–December 2025 finding (1M+ links analysed) was 82%. The work that produces AI citations sits one floor above where most teams are spending their time.

Update your reporting. Dashboards that pull FAQ rich result dimensions through the Search Console API need to be updated before August 2026. If your agency reports include FAQ rich result CTR as a line item, it stops being measurable in two months. Rebuild around AI citation tracking and total page CTR rather than feature-specific snippets.

The bigger pattern — every visibility shortcut has a half-life

What this episode actually shows is not that FAQ schema was a mistake. It shows that visibility shortcuts have half-lives, and the cycle is roughly six years. Feature launches, SEO industry scales it, Google restricts it, Google removes it. Sitelinks, breadcrumb rich results, HowTo rich results, FAQ rich results — the same sequence each time. The pattern says nothing about whether structured data is valuable. It says that any specific extraction shortcut, used as a shortcut, gets withdrawn.

The implications are durable. The work that compounds is the work that does not depend on a specific feature surviving. Entity foundation. Honest content extractability. Editorial citation. Substantive on-page evidence. These do not get withdrawn because they are not features; they are properties of the page itself. The page either contains a defined term inline with a named source or it does not. AI systems either find a corroborated, attributable claim or they do not. Whether Google chooses to render that as a rich result, an AI Overview, an AI Mode follow-up, or nothing at all does not change what is actually on the page.

This is the difference between Floor 2 work that is durable and Floor 2 work that is opportunistic. FAQPage schema as an honest description of a page that genuinely contains Q&A is durable. FAQPage schema scaled across a templated footer because someone wrote a guide about FAQ schema for GEO was opportunistic. The first stays. The second gets removed in May 2026.

For practitioners working on AI visibility specifically, the read is clear. The 92.1% University of Toronto finding and the 82% Muck Rack finding both point to the same place: AI citation is a Floor 3 selection problem, not a Floor 2 retrieval problem, for the queries where AI is making a recommendation. The FAQ schema discussion is occurring on Floor 2. It is structurally one floor below the layer that determines whether you get cited. This does not mean Floor 2 is irrelevant — it means Floor 2 is necessary and not sufficient. Honest Floor 2 work earns you eligibility. Floor 3 earns you the citation. See the AI Citation Gap guide and AEO Is Solving the Wrong 8% for the longer argument.

If the past week has produced anything genuinely useful, it is forcing the conversation up the building. The lift is being built. The temporary ladder has been removed. The floors are still where they were.

Where this fits in the wider picture

Foundational reading: Schema and structured data covers the full schema landscape, including Article, Product, Service, Organization, DefinedTerm, Person, BreadcrumbList. The JSON-LD guide walks through the technical implementation. The schema generator still produces FAQPage markup for the use cases above, with a built-in advisory.

Strategic context: LLM optimisation, the CITATE framework, the four-floor building post. Platform implications: AEO, AIO, Perplexity, ChatGPT, Gemini. If you want a structural read on your own estate before making any of these decisions, the AI Visibility Audit is the entry point.