FAQ Rich Results Are Gone – Why FAQ Schema Stays in My Stack

On May 7, 2026, Google added a banner to the top of the official FAQPage documentation in Search Central. The message: FAQ rich results no longer appear in search results. Period. Not even for government and health sites – the only category that had retained eligibility since August 2023.

The SEO industry reacted on LinkedIn, in newsletters, in Slack channels. Some wrote “killed,” others said “logical endpoint,” others shrugged and said: “We saw this coming.” Christian Kunz from SEO Südwest summed it up dryly: “Sometimes the dead really are dead.”

From my work as Product Developer at iGaming.com and across my SEO Kreativ client projects, I can say: operationally, this changes very little for most of us. Click yields from FAQ rich results have been near zero since August 2023. The more interesting question is: what do you do now with the schema code that’s still sitting in thousands of templates? Strip it out? Leave it? And what do you optimize for instead?

This article gives you the facts unfiltered, a curated selection of industry voices, and my own position – the same one I’m running on seo-kreativ.de and in every client project I touch.

Google didn’t announce the shutdown in a big blog post. They added a notice banner directly at the top of the official FAQPage documentation. The wording, paraphrased: FAQ rich results have stopped appearing in search since May 7, 2026. The FAQ search appearance, rich result report, and support in the Rich Results Test will be removed in June 2026. Search Console API support follows in August 2026 – explicitly framed as a buffer to let teams adjust their API calls.

Three things need to be kept apart:

The last detail tends to get glossed over in most coverage, but it’s the one that matters in practice: you don’t need to strip FAQ schema out of your code just because Google is dropping the rich result. The markup stays valid, other search engines keep processing it, and AI systems use the structured Q&A pairs for their own purposes. Put differently: Google is pulling a SERP feature, not the data spec. Anyone who conflates the two ends up ripping code out of templates in May 2026 that would still have been useful to them in July.

That nuance is the key to everything that follows.

FAQ rich results launched in 2019 and quickly became one of the most aggressively adopted schema types in SEO. The reason was simple: a bit of JSON-LD in the head, and your search result suddenly took up two or three times the SERP pixels. Anyone who got there early and clean took meaningful CTR uplifts in travel, legal, finance, and iGaming. SearchPilot documented this in multiple split tests during the early phase, with organic traffic uplifts between 3 and 8 percent.

The phase didn’t last. Adoption became too fast, too broad, and too sloppy. The timeline of restrictions reads like a sequence of tightening clamps:

• 2019: FAQ rich results go live, JSON-LD via FAQPage schema becomes standard.
• April 2023: First quiet reduction. Search Console shows declining FAQ impressions without an official Google announcement.
• August 2023: Google makes it official. From that point on, FAQ rich results are only displayed for “well-known, authoritative government and health websites.” For everyone else, the feature is practically dead.
• January 2024: SEO expert Brodie Clark observes on X that FAQ rich results disappear even from the remaining sites for periods, briefly come back, then disappear again – Google appears to be testing.
• March 2026: The March 2026 Core Update halves the remaining FAQ rich result impressions according to aggregated tool data – what little was left largely disappears.
• May 7, 2026: Banner in the documentation, FAQ rich results are completely out of the SERP – including the previously privileged gov/health sites.
• June 2026: GSC report and Rich Results Test lose FAQ support.
• August 2026: Search Console API loses FAQ support.

In my work on seo-kreativ.de and in client projects, I’ve documented the descent since mid-2023. For clients with significant FAQ schema deployment, FAQ rich result visibility between August 2023 and March 2026 effectively dropped to zero in several projects. That matches the industry consensus: the feature stopped being a performance factor for the broad mass of sites a long time ago.

Google’s official rationale is sparse. The unofficial one came from John Mueller in 2023, then a Google Search Advocate. Asked why Google was so drastically restricting FAQ rich results, he answered on X (then Twitter), paraphrased: “Independent of abuse and overuse – things can change on the web, with users, or with focus shifts, and it’s important to clean up from time to time. (And it’s a good reminder to read up about the tragedy of the commons.)”

The Tragedy of the Commons is an economics concept: a shared resource gets so heavily exploited through uncoordinated individual self-interest that it becomes worthless to everyone. Translated to FAQ schema: the SEO community collectively trashed the feature between 2019 and 2023. Suddenly almost every product page, blog post, and “About us” page had a FAQPage schema with pseudo-questions. The German SEO blog code78.de captured it in March 2026: “Suddenly almost every product page, every blog article, and every ‘About us’ page had a FAQPage schema sprinkled with pseudo-questions. Is the product good? Yes, our product is very good.”

Behind Google’s official “cleanup,” there are three drivers in my view that the industry largely agrees on:

1. Quality control became impossible. With millions of FAQ-marked pages, manual or algorithmic cleanup wasn’t scalable anymore. Google ended the game instead of refereeing it.
2. SERP simplification as a strategic goal. Google has been systematically dismantling SERP features since 2023 – HowTo rich results in the same breath, later Course Info, Special Announcement, and others. That fits a larger arc: less SERP clutter, more space for AI Overviews.
3. The shift toward an answer engine. Google is replacing the classic SERP with direct answers more and more. If you serve answers directly, you don’t need FAQ dropdowns under the blue links – the answer is already at the top in the AI Overview.

The last point is the key transition to my position below. For more on how AI Overviews work technically and which signals actually count there, I broke it down in Google AI Overviews: How It Works Explained Simply.

I reviewed reactions on LinkedIn, X, Bluesky, and the German-language SEO blogs in the days before and after May 7, 2026. Three currents separate clearly.

The pragmatists: “Finally gone”

Akash Sehgal pointed it directly on LinkedIn right after the announcement: “BREAKING: GOOGLE just KILLED FAQ rich results. SEOs for the last 3 years: ‘add FAQ schema everywhere, get more SERP space, increase CTR.’ Google in 2026: ‘yeah… we’re done with that.'” His advice: “Stop adding FAQs just for schema. Start answering real questions inside your content.” This faction says, in essence: anyone who optimized for the pixel game has long since shot their bolt – schema work should follow real user questions, not SERP appearance.

The schema hygienists: “Leave it alone”

Rich Missey wrote on LinkedIn in 2023 – and his position has lost none of its currency in 2026: “Don’t make Googlebot think. The easier we make it for the bots/systems/algorithms to understand the content of a URL and the context in which that content exists, the better our chances.” This faction says: schema stays in, because removing costs effort and keeping costs nothing.

The grounding faction (DE): the most differentiated position

This is where it gets interesting. Three days before the final shutdown, on May 4, 2026, Olaf Kopp (Aufgesang) published an article on aufgesang.de – one of the strongest German analyses on the topic in 2026. Olaf was a long-time schema skeptic and explicitly relativizes his position now. His key sentence for our subject: “Thanks to FAQ markup, search engines were able to learn how to answer questions. This newly acquired ability now serves as a foundational basis for AI Overviews, AI Mode, Gemini.” In other words: as an industry, we collectively taught Google what a question and an answer are. That capability now sits in the models.

Olaf’s overall verdict: “Strategically I would view Schema.org as a ‘hygiene factor’ for search engines and as a useful indirect lever for GEO/AI search, not as a standalone ‘LLM lever’.” That’s a markedly more cautious framing than the widely cited but empirically unverified “3.2x citations” claims from the English-language scene – and a particularly sober and differentiated framing in the German-speaking SEO scene.

Hanns Kronenberg (Head of SEO at Chefkoch, previously Sistrix Head of Product) goes one step further. In 2025 he launched the Grounding Page Standard – an open format for brands to provide core facts in machine-readable form for AI systems. Kronenberg’s conceptual pair: “On-Model SEO” (a brand’s representation in the internal model knowledge) versus “Off-Model SEO” (external referenceability and retrieval capability). FAQ schema sits squarely in Off-Model SEO under this logic – as part of a domain’s machine-readable identity layer. That’s exactly the lens through which I look at structured data in 2026.

Markus Hövener (Bloofusion, suchradar, Search Camp Podcast) had already framed the original 2023 reduction on LinkedIn – pragmatically and without hype. In May 2026 his online seminar “SEO Deep Dive: Everything You Need to Know About Markup!” remains in his program. The position behind it: structured data stays relevant craft, but FAQ was always the over-stretched case. Whoever implemented the markup cleanly leaves it in place. Whoever used it as a pixel hack now removes the generic blocks. Christian Kunz of SEO Südwest brings the DACH position to a single line – see the intro: “Sometimes the dead really are dead.”

Internationally, Chris Long (VP Marketing, Go Fish Digital) argues along similar lines. In June 2025 he shared on LinkedIn an analysis on the semantic relevance of FAQs for AI search with the hook: “SEOs, using FAQs in your content could be a GREAT way to optimize for AI Search.” Long is explicitly not in the schema-pragmatist camp; he sees FAQs as a useful format for findability through AI systems.

The international position: cautiously optimistic

John Mueller (Google Search Advocate) has positioned himself on the question multiple times in 2025 and 2026. In November 2025 he replied on Reddit to the worry that Google would abandon schema entirely: “Markup types come and go, but a precious few you should hold on to, like title and meta robots.” On the specific question of whether schema helps LLMs understand entities, he wrote in early 2026: “The short answer is yes, no, and it depends. Some features thrive with structured data… Other features could theoretically be understood from a page’s text, but it’s just so much easier for machines to read machine-readable data instead of trying to understand your page.” So Mueller is neither advocate nor skeptic – he says: for machines, markup is easier to read than running text.

Lily Ray (VP SEO Strategy at Amsive, founder of Algorythmic) emphasizes in her Affiliate Summit West 2026 keynote that strong technical SEO, structured content, and authoritative signals remain foundational to AI search visibility – without canceling FAQ schema, but also without selling it as a cure-all. Aleyda Solis (Orainti, SEOFOMO) puts it succinctly in her AI Search Optimization Roadmap: “Use structured data to help AI models better classify and extract structured answers.” Both confirm: schema is part of the playbook, not the decisive lever.

The “3.2x citations” claim – and why I’m careful

Persistent claims circulate in the English-language scene that pages with FAQ schema have a “3.2x higher probability of appearing in Google AI Overviews” or that “72% of cited content includes structured Q&A format.” You’ll find these numbers on portalZINE.DE, faqjsonld.com, roardigital.co.uk, and dozens of other posts. I have not been able to find the methodology of these studies in several hours of research – neither sample size nor comparison group nor the definition of “citation.” Anyone who hits you in 2026 with hard multipliers should be able to link the study. So far nobody does.

My own roll-out: FAQ schema during the shutdown

On seo-kreativ.de I’ve been rolling out FAQ schema retroactively to existing blog posts in the weeks before and after May 7, 2026 – precisely in the window where the SEO community is ripping it out of templates. That’s deliberate.

As of May 11, 2026: 26 valid FAQPage elements in Google Search Console. The date cluster in the GSC report shows three phases: roll-out start mid-March with two to three test elements, steady-state until April 23, then a scaling wave from 9 to 22 elements within five days.

My goal: Every post and page on seo-kreativ.de with a real FAQ section – meaning helpful user questions, not bloated pseudo-questions – gets correct FAQPage schema. Full coverage as the end state. The roll-out continues, and I expect further jumps in the GSC count over the coming weeks.

The reason isn’t Google, it’s AI search. Structured data is a hygiene factor for LLMs – Olaf Kopp calls it that, and I agree. Anyone ripping FAQ schema out of their templates in the week after May 7, 2026 is optimizing for yesterday’s SERP feature. Anyone curating and building it up is optimizing for the answer machines of the next 24 months.

I’m convinced structured data is useful for LLMs – not as a magic citation lever, but as a clean Q&A format that any embedding pipeline, any Graph-RAG index, and any future crawler processes without friction. If that assumption is right, the roll-out costs me almost nothing today and delivers differentiation in 12 to 24 months. If it’s wrong, I’ve at least documented my FAQ inventory in a structured way – that has operational value too.

In my client projects – from Swiss PropTech to solar to e-commerce – I’ve stopped using FAQ schema primarily to win SERP pixels since mid-2024. That was never the goal anyway, because the game was already over in August 2023. My goal is grounding: AI systems like Google AI Overviews, Perplexity, ChatGPT Search, Gemini, Claude, Bing Copilot, and the answer engines that will join in 12 or 24 months should recognize my content as a reliable, unambiguously structured source.

Structured data is the technical lever that gives machine readers a second, unambiguous layer on top of running text. For more on how that works in detail and why it specifically matters for AI Overviews, see Structured Data and AI Overviews.

Here’s the point most 2026 discussions miss – and which Olaf Kopp soberly highlights in his Aufgesang piece: schema markup isn’t only a live signal for the current AI generation. It’s also a training signal for the next one. Every crawl, every index entry, every Wikidata sync that runs today with clean FAQPage schema feeds into the training data of tomorrow’s models. Olaf’s example illustrates it well: Google urged e-commerce sites for years to add product schema in shopping feeds – today Google can recognize these attribute-value patterns even in unstructured content, because the model learned from the tagging patterns. Exactly the same has happened with FAQPage: as an industry, we collectively taught the models what a question and an answer are.

My four preconditions for FAQ schema to even be used by me or for clients are strict:

1. The questions are real user questions. Source: Search Console (queries, position 5–20), People Also Ask, Reddit threads, support tickets, sales calls. No invented questions, no “Is our product good?”.
2. The answers stand on their own. Each answer must work when ripped out of context – that’s exactly what AI systems need for extraction. 40–80 words, no marketing fluff, with concrete numbers or facts.
3. The FAQ section isn’t a mandatory block. If a topic doesn’t have honest, open user questions, there’s no FAQ. A FAQ block on every category page of a shopping site is exactly what Google killed in 2023. I don’t play that.
4. The markup mirrors only what’s visible on the page. Hidden FAQ content that only exists in the JSON-LD is clearly against Google’s guidelines. Even without a rich result, that can do damage.

If those four preconditions are met, the schema stays. Reason in one sentence: it costs nothing, doesn’t hurt, and we’re building long-term grounding structures for an answer-driven search world whose final form we don’t yet know.

Important framing – and here I follow Olaf Kopp’s and John Mueller’s sober position: a direct, causal link between FAQ schema and the citation probability in Google AI Overviews or Perplexity is not backed by an official statement from Google or OpenAI. What’s documented: structured data reduces ambiguity, helps crawlers and LLM-based systems with parsing, feeds into Knowledge Graphs, and is explicitly used in Graph-RAG architectures. John Mueller himself put it in the simplest formula on Reddit in early 2026: “It’s just so much easier for machines to read machine-readable data instead of trying to understand your page.”

Here’s where almost every 2026 article on the topic gets shallow. Most authors list “other search engines” and “voice search”, tack on a reference to AI search, and call it done. I go a step further – because from my work as Product Developer at iGaming.com and from my own tooling work (semantic crawlers, vector databases) I know exactly where structured data is actually connectable outside Google Search. Here are six concrete paths, ordered by decreasing verifiability – with a clear judgment on what’s active today and what’s a longer-term bet.

1. Other search engines keep parsing schema

Bing, DuckDuckGo (which sits on the Bing index), and Brave Search use FAQPage markup unchanged. Fabrice Canel, Principal Product Manager at Microsoft Bing, explicitly confirmed in March 2025 that structured data helps Microsoft’s LLMs understand content for Copilot – one of the few official vendor statements we have on the topic in 2026. Bing’s market share in Germany may be small, but keeping the markup costs zero.

2. Knowledge Graphs and Wikidata anchoring

FAQPage schema with correctly linked about or mentions properties pointing to Wikidata IDs helps anchor entities in knowledge graphs. Hanns Kronenberg developed a dedicated vocabulary for this with his Grounding Page Standard. FAQ structures don’t just deliver facts – they deliver the relationship between question and answer, i.e. semantic edges that a knowledge graph can store as triples (entity – property – value).

3. Graph-RAG architectures in the AI world

Olaf Kopp describes this precisely in his Aufgesang piece: in Graph-RAG systems, schema markup is explicitly turned into nodes (entities) and edges (relationships). The LLM then receives not the raw JSON-LD block, but pre-processed facts – tabular or in natural language – extracted from the markup. FAQPage schema sits right in the grounding layer here.

4. Training signal for the next model generation

Even if LLMs see JSON-LD only as a token sequence during pre-training (Olaf Kopp rightly emphasizes this point), the model learns from the pattern: “A block with @type Question, followed by a block with @type Answer, is a FAQ structure.” Every additional FAQ page with clean schema on the web increases the training sample for future models. That’s an indirect but long-term effective bet.

5. Voice search and smart assistants

Alexa, Google Assistant, and Siri (with the deeper Gemini integration expected from 2026) extract answers from Q&A-structured sources with higher probability than from running text. Notably, Google Assistant Actions are still explicitly listed as a use case in Google’s FAQ documentation – even after the May 7, 2026 update. Voice isn’t a 2026 SEO gold mine, but a non-zero channel that benefits from clean FAQ markup.

6. The path almost nobody sees: your own RAG pipeline

Here’s where it gets really interesting, and where the 2026 discussion has a hole. If you’re building a support-bot system inside your company, maintaining an internal knowledge base with semantic search, or developing a product-internal AI layer (RAG with ChromaDB, Pinecone, Weaviate, or pgvector), you need exactly one thing: atomic, clearly bounded Q&A units in a standardized format. FAQPage schema has been precisely that format since 2018. Per question, it gives you a separate @type Question block with name (question) and acceptedAnswer.text (answer). That is – speaking purely technically – an embedding-ready data feed you don’t have to teach anything more to.

In my own projects I read FAQPage JSON-LD out of the CMS, push the individual Q&A pairs through a sentence-transformers model, store the vectors in a ChromaDB collection, and have a semantic search layer over my FAQ inventory within minutes. Without schema, I’d first have to extract Q&A from HTML, parse it, segment it – with schema, it’s a JSON.parse(). From this perspective, FAQ schema in 2026 is what XML sitemaps were for SEO in 2010: a data structure you shouldn’t actually be building for Google but for your own infrastructure. And for every AI layer that gets bolted on top in the coming years – internal or external.

So what to actually do? Four steps, starting today

If I translate these six paths into a concrete recommendation – i.e. what you should operationally do when you close this article – it boils down to four steps. I do all of them in my own projects and with clients. They cost half a day combined.

1. Treat FAQ content as a vector-database source from now on. Write every answer so that it works out of context (40–80 words), states a clear claim (no “it depends”), and is referenceable by concrete numbers or terms. That’s good for AI citations, good for voice, good for your own RAG layer.
2. Enrich FAQPage with entities. In every Q&A block, set an about property pointing to a Wikidata ID, Wikipedia URL, or your own sameAs collection. That way every knowledge-graph builder, every Graph-RAG layer, and every future model knows what you’re talking about. Without entity anchors, FAQ schema is half-deaf.
3. Version your FAQ answers. Set dateModified in every entry, ideally per Question. Freshness is a verifiable signal in nearly every AI pipeline in 2026, and it protects you from the “stale schema” effect Stackmatix describes in its 2026 schema guide: AI systems that see outdated schema data reduce trust in all your content – not just the old.
4. Make your FAQs machine-accessible – not just via schema. Additionally export your FAQ inventory as JSON feed, RSS, or llms.txt-conformant format. Yes, John Mueller called llms.txt overrated in 2025, and for most use cases he’s right. But for steering your own RAG pipelines, for internal knowledge-base sync, and for Bing/Brave/Perplexity bots that read exactly these helper files, it’s a zero-cost lever with non-zero impact.

These four steps combine into a position that’s still rarely articulated in 2026 mainstream SEO discourse: FAQ schema is no longer an SEO markup. It’s a Q&A API contract between your CMS and any AI layer – internal or external. Anyone who thinks of it that way doesn’t panic on May 7, 2026, doesn’t rip the markup out of templates, and instead keeps building out the data structure they already have.

When a client calls me after May 7, 2026 with “Do we have to remove this from 800 category pages now?”, my standard answer is: audit honestly first, then decide. The SearchPilot test from late 2024 – statistically clean – showed that removing FAQ schema produced no statistically significant organic traffic effect. The SEO world doesn’t end if you leave it. The real question is: is the cost of removing smaller than the potential value of keeping it?

Three scenarios from my practice:

Scenario A: Schema garbage – out it goes

Symptoms: generic, identical FAQ blocks on 200+ pages, often copied from a template. Questions like “Does your shop also offer [Category X]?” with the answer “Yes, we offer [Category X].” No real search intent covered, no value for users. These blocks are listed in Google’s documentation since 2023 as examples of “non-permitted FAQ usage.” Even without a rich result, they can act as a negative quality signal. My call: remove the block, strip the markup, invest the energy in real content.

Scenario B: Real FAQs – leave them

Symptoms: a handful of specific questions per page, each with an individually written answer that users actually ask. Keywords match GSC queries or People Also Ask. Markup mirrors visible content. That’s schema you keep. Maintenance cost is near zero, AI systems parse it, and the moment Google announces a different way of using FAQPage data (e.g., an AI Overview module), you’re already there.

Scenario C: Templated with maintenance debt – curate

Symptoms: an FAQ section that started sensibly and hasn’t been touched in two years. Some answers are outdated, some questions don’t fit the current product anymore. The more expensive but correct path: walk through, halve, sharpen, update. What’s left is usually better than what was there before – and falls into Scenario B.

None of these scenarios calls for panic. A blanket sitewide schema removal “because Google doesn’t use it anymore” is just as ill-considered as the templated FAQ block from 2021.

What I’m doing for seo-kreativ.de and most client projects in the coming weeks – in this order:

1. Export historical FAQ GSC data. In Search Console under “Enhancements” → “FAQ”, pull all data from the last 16 months as CSV. Before June 2026 – after that, the filter is gone.
2. Identify affected pages. Which pages have had FAQ rich result impressions in the last 12 months? Hold those in a list – you’ll want to monitor them more closely for 6–12 weeks to separate real traffic/CTR effects from market noise.
3. Update reporting and dashboards. If you have Search Console API pulls on the FAQ dimension, plan the rebuild by August 2026. In client reports, I’d stop showing FAQ clicks and impressions as a separate metric immediately.
4. Run a schema audit. Following the three scenarios above (garbage / real / templated). For larger sites, a Screaming Frog crawl or a custom tool that extracts all JSON-LD blocks and filters by FAQPage is worthwhile.
5. Switch the Rich Results Test workflow. No longer usable for FAQ validation as of June 2026. Alternative: Schema.org Validator for pure syntax checks. Still useful for the schema markup itself.
6. Rethink the FAQ strategy. If you keep FAQ schema strategically, do it not for Google pixels but for grounding and AI citation probability. For how that fits into the broader AI search strategy, see Google AI Mode vs. AI Overviews for the structural differences and what they mean for SEOs.

From my work in technical SEO audits, I still see regularly that FAQ sections in client projects sit as “standard blocks” on every category page, with no one having looked at them in months. Those blocks are candidate number one for Scenario A. If you were planning to clean up in 2026 anyway, this is the moment.

May 7, 2026 isn’t an earthquake. It’s the formal closure of a process that started in August 2023. The operational impact on most sites’ traffic is near zero – the game was already over. The strategic impact still matters: FAQ schema leaves the list of “SEO tricks for more pixels” and moves to the list of “grounding signals for AI search.” Those two categories have different requirements, different success metrics, and different risks.

My recommendation – short and undiplomatic: keep FAQ schema where it earns its place. Clean up where it was just standard templates without substance. Invest the energy you’ve spent on pixel optimization for years into real answers to real questions. That works in classic search, in AI Overviews, in Perplexity, and in any future answer engine we don’t know yet.

If you want to dig deeper into the structural shift in the SERP and what it means for your optimization, I recommend Zero-Click Search 2026: Data, AI Overviews & What to Do. It covers the Sistrix and SparkToro data on the DACH reality and lays out concrete moves for the post-click world.