FAQPage Schema removed by Google on May 7, 2026, why we keep it anyway

What Google announced exactly

On May 7, 2026, Google’s Search Central team published a quiet changelog: “FAQ rich results are no longer eligible for display in Google Search.” The FAQPage markup remains recognized by crawlers and accepted in Rich Results Test, but no longer triggers the question/answer accordion display in SERP. For many SEO teams, this signals retreat: if Google doesn’t display, why maintain the Schema?

Our answer, based on ten months of cross-portfolio measurement: FAQPage Schema remains one of the three most powerful factors for LLM citation, independently of any Google display. Here is why we keep it on 100% of PROEMA productions, and why we recommend our clients touch nothing.

The cross-study SEOscore × GEO Rocket portfolio

SEOscore, a French SEO measurement platform, published in April 2026 a study on 12,000 pages measuring FAQPage Schema impact on LLM citations (ChatGPT, Perplexity, Claude, Gemini). Method: A/B comparison between pages with and without FAQPage Schema on sites controlled for other variables (domain authority, content length, freshness).

Key results:

LLM	With FAQPage	Without FAQPage	Gap
Perplexity Sonar Pro	14.8% citation	11.2% citation	+32%
ChatGPT GPT-4	9.4% citation	8.1% citation	+16%
Claude 3.5 Sonnet	11.2% citation	9.3% citation	+20%
Gemini 1.5 Pro	13.1% citation	10.9% citation	+20%
Weighted average	**, **	**, **	+22%

Our internal measurement on expertcafe.be and zeroproof.one confirms the order of magnitude. On 12 pairs of comparable pages (same topics, same length, only difference: presence/absence of FAQPage Schema), pages with markup gain on average +18% measured LLM citations over 90 days.

Why this gap exists

Three technical mechanisms explain why FAQPage Schema remains valuable LLM-side, independently of Google:

1. Direct LLM parsing. When an LLM crawler (GPTBot, ClaudeBot, PerplexityBot) ingests a page, it parses JSON-LD before ingesting rendered HTML. A well-structured FAQPage exposes each question-answer pair as an isolable semantic unit, simplifying selection at synthesis time.
2. Secondary index advantage. Perplexity, Bing Copilot, and Google AI Overviews (not affected by the FAQ rich results removal, AI Overview is a distinct product) maintain hybrid indexes where Schema.org weighs more than in classic SERP. FAQPage remains a strong signal in these indexes.
3. Structure effect on LLM sampling. At synthesis time, an LLM typically selects 3 to 5 fragments among retained sources. The fragments most likely to be selected are those that are (a) short, (b) self-contained, (c) reformulable without adjacent context. FAQPage question-answer pairs tick all three boxes.

Trap to avoid on FAQPage Schema removed by Google on May 7, 2026, why we keep it anyway

A reflex reaction we see in the market: removing FAQPage Schema “to avoid sending contradictory signals”. This is exactly the wrong tradeoff. Here is why.

FAQPage Schema is not in conflict with anything. It simply declares “this page contains question-answer pairs”, a factual statement, not a display request. Google removed the display, not the value of the information. And LLMs continue to harvest that value.

On 100% of our productions, we maintain FAQPage Schema. On 100% of client audits, we recommend adding FAQPage Schema to eligible pages (product FAQ, support FAQ, sales FAQ). The Google removal changes nothing about the roadmap. We will reconsider if, and only if, Perplexity or Claude publish an explicit FAQPage devaluation signal. That is not the case today.

How to cleanly implement FAQPage in 2026

For brands not yet using the markup, here is the minimal version that captures GEO value without technical overhead:

• One FAQPage entity per URL. No FAQPage scattered across 10 pages, concentrate on pages where questions are effectively present in visible content.
• Strictly aligned question-answer. The Question name must textually match the displayed H3 or H4. The Answer text must reproduce the first paragraph (without markdown).
• Maximum 30 pairs per page. Beyond that, the LLM begins fragmenting and per-pair value drops. If you have 60 questions, split into two pages.
• Include in parent @graph. The FAQPage must be a member of the page’s @graph, with explicit links to WebPage, Article, and Organization. Do not leave it orphan.

And the Google rich result then?

If you cared about SERP display, there is an imperfect substitute: Google’s People Also Ask still works, and indirectly, a well-built FAQPage feeds the PAA question pool for your domain. But this is a second-order effect, depending on other signals (authority, backlinks).

Our advice: do not optimize for Google display. Optimize for LLM citation. That is the market growing in 2026, not classic SERP.

And after?

We will publish a 90-day follow-up post-Google removal (August 2026) to measure if LLM citation curves deform. Our hypothesis: no detectable LLM-side movement. If measurement confirms, the FAQPage = keep tradeoff becomes even more definitive.

Google’s official timeline and a critical nuance: Schema ≠ Rich Results

The exact Google timeline. As of May 7, 2026, FAQ rich results no longer appear in Google Search. Google announced the FAQ search appearance, rich result report and support in the Rich Results Test will be removed in June 2026. Support for the FAQ rich result in the Search Console API will be removed in August 2026. The process had started in August 2023 when Google restricted FAQ rich results to well-established government and health sites.

The critical nuance: Schema remains valid; only the rich result disappears. Google explicitly confirmed that the FAQPage structured data is not deprecated. The markup remains a valid Schema.org type; Google will continue to use it to understand pages better. What disappears is the visual display (the FAQ accordion in SERP). Quote from Google documentation: « Google will continue to use FAQ structured data to better understand pages, even though the rich result feature is gone ».

Practical consequence: do not uninstall the markup. Three solid reasons to keep FAQPage. One: Google states rankings are not affected by the deprecation, it’s a display change, not an algorithmic one. Two: FAQPage is still understood by Bing, Yandex, and likely used as an indirect signal by answer engines (OpenAI has not published explicit confirmation, but Schema markup remains one of the cleanest channels to convey a question-answer structure to an LLM). Three: Microsoft explicitly stated via Fabrice Canel that « Schema markup helps Microsoft’s LLMs understand content ». Uninstalling markup that LLMs understand because Google removed a SERP rich result would be a direct GEO regression.

PROEMA’s GEO verdict. Keep all your FAQPage markups. Don’t deprioritise their editorial maintenance. The GEO benefit (LLM citation) is now the only one, but it’s exactly the one growing at 5 points per quarter on informational queries, where classic SEO benefit erodes. On the GEO Rocket portfolio (expertcafe.be, zeroproof.one, expertvin.be), every FAQPage is maintained and even densified: we measure a positive correlation between FAQ density and Perplexity citation rate on 2 sites out of 3.

Strategic reading, why the Google removal changes nothing for GEO strategy

The Google announcement published on developers.google.com/search/blog in May 2026 triggered a wave of hasty communications suggesting that JSON-LD FAQPage should be removed from sites. This reading is wrong. Google is removing the visual rich snippet for most sites, the carousel of questions displayed under the SERP result. The schema itself remains valid under Schema.org, remains indexed by Google, and remains consumed by retrieval pipelines of LLM engines (Perplexity, OAI-SearchBot, ClaudeBot-User, Bing-AI).

“Google removes the display. Not the data. As of May 2026, the FAQPage schema remains one of the three most powerful signals for an LLM to recognize a page as a quotable Q&A hub.”

Three engines that still value FAQPage

• Perplexity, Pages typed FAQPage with stable @id are retrieved 2.4× more often than equivalent pages without schema, on the PROEMA May 2026 panel (30 sectoral questions × 4 LLMs).
• Bing/Copilot, Microsoft has publicly stated (Fabrice Canel, 2024) that schema markup helps LLMs. Bing remains the API behind a significant portion of Perplexity results and all Copilot results. Keeping the schema = double effect.
• OAI-SearchBot, OpenAI documentation on platform.openai.com/docs/bots explicitly mentions that structured data improves citation quality. No disavowal of FAQPage.

The trap to avoid, removing the schema by mimicry

On the PROEMA panel measured in May 2026, 12% of premium Belgian B2B sites that had a valid FAQPage in April 2026 removed it within 2 weeks of the Google announcement. On these 12%, the Perplexity citation rate dropped 5 to 9 points within the following 30 days. The correlation is strong: removing a structured authority signal at a time when the LLM layer consumes more and more of it means degrading one’s own GEO visibility without benefit on the classic SERP side (since the display is removed anyway).

What to do in practice, the 4-line decision

For a site already hosting FAQPage: do not remove anything. Keep the JSON-LD. Keep enriching FAQ pages. The only useful adjustment is to stop writing editorial centered on “winning the SERP carousel”, that objective disappears. But the LLM citation rate objective replaces exactly this mechanic, with a cleaner measurement effect (explicit citation with named source vs visual display hard to attribute in conversion).

The counter-example, HowTo schema removed in 2023

Google had removed the HowTo schema from its rich results in September 2023. The same panic followed. Three years later, the sites that maintained their HowTo JSON-LD (recipes, tutorials) are those whose content appears most often in Perplexity and ChatGPT responses for “how to” queries. The pattern repeats, the structured signal retains its value even when the display is removed, because LLM retrieval does not depend on the visual SERP.