Schema · Primary sources · May 2026
FAQ schema is dead. The data says so. Twice.
On May 7, 2026 Google removed FAQ rich results from Search. Two weeks earlier, the first causal study on schema and AI citations found no measurable boost. This is what the primary sources actually say — and what we're changing in That SEO Agent.
Chapter 01
What Google actually changed on May 7, 2026
Google's official documentation on FAQPage structured data now opens with a single sentence that ends a decade of FAQ-driven SEO: "As of May 7, 2026, FAQ rich results are no longer appearing in Google Search." This is not a restriction, not a gradual rollout — it is the end state.
The timeline matters because tooling is being decommissioned in stages. FAQ rich results stopped appearing immediately. The Rich Results Test will drop FAQ support in June 2026. The Search Console API will follow in August 2026. If your dashboards still chart FAQ impressions, they will go dark by Q3.
There is one carveout: FAQ rich results are still served for "well-known, authoritative websites that are government-focused or health-focused." For everyone else — SaaS, e-commerce, media, agencies — the schema produces nothing in Google.
This is the second time Google has killed FAQ. In August 2023 they restricted the feature to gov/health sites. In May 2026 they finished the job. The pattern matters more than the date, and we'll come back to that in Chapter 04.
| Milestone | Date |
|---|---|
| First restriction — gov/health only— Soft deprecation | Aug 8, 2023 |
| FAQ rich results removed from SERP— Hard deprecation | May 7, 2026 |
| Rich Results Test drops FAQ support | June 2026 |
| Search Console API drops FAQ data— Dashboards go dark | August 2026 |
| FAQ rich results in Bing— Oct 2025 guidance | Still supported |
Chapter 02
The Ahrefs study: schema does not boost AI citations
While the SEO community was debating whether to add more schema for AI search, Ahrefs ran the only thing the conversation was missing: an actual controlled experiment. Published in 2026, it tracked 1,885 pages that added JSON-LD schema between August 2025 and March 2026, matched against 4,000 control pages with similar pre-treatment citation levels.
The headline result: adding schema produced no major lift in AI citations on AI Mode (+2.4%, not significant) or ChatGPT (+2.2%, not significant). On Google AI Overviews it produced a statistically significant decline of −4.6%. This is the inverse of what most SEO blogs were claiming.
The methodology matters because the original correlation was strong. Pages cited by AI were almost three times more likely to have JSON-LD than non-cited pages. But when Ahrefs controlled for confounders — technically sophisticated sites tend to have both schema and citations — the causal effect collapsed.
| Surface | Citation change after schema | Significance |
|---|---|---|
| Google AI Overviews | −4.6% | Significant |
| Google AI Mode | +2.4% | Not significant |
| ChatGPT Search | +2.2% | Not significant |
Chapter 03
What Google's AI Optimization Guide actually says
In December 2025 Google published its first official guide for optimizing content for AI features in Search. It is the closest thing to a primary source for what Google AI Overviews and AI Mode reward. The document is short and unambiguous about what does not work.
Google explicitly lists four things you can ignore: chunking content into tiny pieces, rewriting content specifically for AI systems, creating AI-specific files (this is widely read as including llms.txt), and pursuing inauthentic mentions across the web. These are not soft suggestions — the document tells site owners to "ignore" them.
What Google says does work is the same thing it has said for years: be indexable, follow snippet eligibility rules, comply with spam policies, and create "unique, compelling, and useful" content with original perspective. Structured data "remains a good idea as part of your overall SEO strategy" but "isn't required specifically for AI features."
The implication is uncomfortable for the GEO industry: there is no special schema, no markup, no file, no incantation that gets your content into AI answers. The mechanism is the same content quality bar Google has asked for since at least 2022 — just harder to game.
| Tactic | Google's stance | Industry consensus (often wrong) |
|---|---|---|
| Add llms.txt | Not required, not supported | "Add it just in case" |
| Chunk content for retrieval | Can be ignored | "Smaller chunks help RAG" |
| Rewrite copy for AI | Can be ignored | "Write for AI first" |
| Add FAQPage schema | Optional, no AI feature benefit | "Most cited format" |
| Be indexable, satisfy users | Required, primary signal | Underrated |
Chapter 04
Three myths the data already disproved
Chapter 05
What to change this week
Chapter 06
The cycle, and why it will repeat
Joost de Valk's central observation is the most useful frame for the next two years: anything that can be spammed in SEO will be spammed, and platforms eventually withdraw the reward. The FAQ schema arc — introduced, abused, restricted, killed — is happening again with AI optimization claims that have no primary source. The defense is the same as it ever was.
Run a schema-honesty audit
Stop optimizing for rich results that no longer exist.
That SEO Agent's GEO and AI Visibility analyzers are being recalibrated this quarter so your score reflects what AI engines actually reward — not what the 2023 SEO blog posts claim.
Free during beta · Primary sources only