Google killed FAQ rich snippets. Here's what to do.

What Google actually changed

In Google's own words: "As of May 7, 2026, FAQ rich results are no longer appearing in Google Search." The Rich Results Test — the tool a generation of SEOs leaned on to validate their schema — is being retired in June 2026.

This wasn't a sudden ambush. Google quietly restricted FAQ rich results to health-focused and government sites in August 2023. For three years most SEO teams kept the JSON-LD on their pages out of habit and a quiet hope. Now Google has flipped the last switch.

Why your FAQ schema still earns its keep

The thing every "is FAQ schema dead?" hot-take is missing: Google is one consumer of structured data, not the only one. Here's who still reads your FAQPage JSON-LD and acts on it:

• ChatGPT's web tool uses page schema to identify Q&A structure and decide which answers are quotable. Schema-marked FAQs get cited at noticeably higher rates than the same content in raw paragraphs.
• Perplexity relies heavily on structured data for source attribution — the answer cards you see citing a specific Q frequently come from a FAQPage block.
• Gemini and AI Overviews extract entities from JSON-LD when building their answer summaries. The schema is the cleanest signal that a page is structured around questions.
• Bing still renders FAQ rich snippets in its SERP. Google quit; Microsoft didn't.

The value didn't disappear — it migrated. Where a green checkmark in Rich Results Test used to translate to extra real-estate on a Google SERP, the same schema now translates to a citation in an AI assistant's answer. The currency changed; the engraving on the coin didn't.

Why "the SERP is dead" undersells what actually happened

Some commentary frames the May 2026 announcement as the end of FAQ-as-SEO-tactic. That overshoots. What's actually happened over the last 18 months is a quieter handoff: when someone asks Google "how do I X?", more and more of those queries never produce a click. Either an AI Overview answers in-line, or the user starts in ChatGPT, Gemini, or Perplexity to begin with — and the assistant cites a couple of sources at the bottom of its reply.

The traffic that used to go through SERPs is now intermediated by an assistant. Your FAQ either gets cited (and gets referral traffic plus the harder-to-measure brand surface) or it doesn't. The schema is still how the assistant identifies you as having a clear, citeable answer.

What to actually audit now

"FAQ rich snippet validation" was a checklist with three or four items. AI citability is a richer surface. The signals AI engines pick up on for FAQ content fall into three buckets:

1. Schema validity — Is there a real FAQPage block? Does mainEntity have 2+ Question entities? Does each have acceptedAnswer.text that matches the rendered DOM? inLanguage set for non-English? Schema/DOM drift reads to AI engines as a cloaking signal and gets you penalized in citation ranking.
2. Citability per engine — Each engine has different preferences. ChatGPT likes 40–60-word lead answers with the conclusion up top. Perplexity loves explicit numbers and dates inside the answer. AI Overviews truncates aggressively; if your answer buries the lede past character 200 it never gets quoted. Gemini favors complete sentences over fragments. The score for a given Q&A should reflect each engine's preferences, not a single average.
3. Voice fit — Is the answer self-contained, or does it depend on prior context the engine won't have? Is the tone helpful or marketing-y? AI engines actively avoid quoting answers that read like ad copy.

The audit work isn't harder than it was for Google rich results — it's just structured around a different question. Instead of "will this render a rich card?", the question is "would an LLM choose to quote this paragraph?"

Concrete next steps

1. Keep the FAQPage JSON-LD on your pages. Removing it because Google won't render it anymore is throwing away the signal AI engines rely on. The schema costs you nothing to keep.
2. Stop treating Rich Results Test as a finish line. It's being deleted in June. A green checkmark there was always a low bar — it tested validity, not quotability.
3. Audit your FAQs against AI engine criteria. That's roughly: answer length 40–80 words with the lede first, question phrased like a real user query (not a SEO header), schema text matches DOM, no marketing voice. Do it manually or use a tool.
4. Watch your AI referral traffic, not just your Google clicks. ChatGPT, Perplexity, and Gemini send referral traffic with identifiable referrers. If your AI-referred sessions are flat while your Google sessions decline, the problem is upstream — your FAQ isn't getting cited.

The thing nobody's saying out loud

Google's FAQ rich snippet was always a participation trophy. Sites that earned the rendering got a tiny SERP bump; sites that didn't lost nothing important. The fact that Google retired it in 2026 — quietly, without much industry pushback — confirms it wasn't material to most outcomes anyway.

What is material is whether your answers get picked when a real user asks an LLM a real question. That's measurable now, optimizable now, and predictably going to matter more, not less, over the next five years. The work hasn't ended; it's just moved.