TL;DR
- Google removed FAQ rich results from search entirely on May 7, 2026, with no announcement beyond a developer doc update. The feature had already been restricted to government and health sites since August 2023.
- Google's own platform guidance now says structured data isn't required for AI Overviews or AI Mode — there's no special schema.org markup for generative results.
- FAQPage schema still has a job: it's a clean, machine-parseable Q&A format that LLM retrieval pipelines can lift directly, even though Google no longer renders it visually.
- Keep FAQ schema on pages where the markup mirrors content that's already genuinely useful to a human reader — don't keep it as a standalone SEO trick with no visible value.
- The real lever for AI citation isn't the schema tag, it's whether your Q&A pairs match actual phrasing people type into ChatGPT, Perplexity, or Gemini.
FAQ schema (schema.org's FAQPage type) is structured markup that wraps a list of questions and answers in machine-readable JSON-LD. It was built so search engines could display expandable Q&A snippets directly in results. That visual payoff from Google is now gone for essentially every site. The question worth answering isn't "does Google still show this," it's "does marking up Q&A content still help a brand get surfaced inside an AI-generated answer."
What actually changed, and when
Google's rollback happened in two stages. In August 2023, the company restricted FAQ rich results to a narrow set of "well-known, authoritative" government and health domains — everyone else lost the visual snippet overnight, even though the schema still validated fine in testing tools. Most site owners noticed the click-through drop before they noticed the policy change.
The second stage landed on May 7, 2026: Google removed FAQ rich results from search entirely, for every site, including the government and health exception. There was no blog post, no changelog entry with fanfare — just an updated note buried in developer documentation. The deprecation is now working its way through the full stack: FAQ enhancement reports are disappearing from Search Console, FAQ support in the Rich Results Test ends in June 2026, and Search Console API support for the FAQ report ends in August 2026. Functionally, Google is unwinding every tool that let you monitor or validate FAQ markup, not just the rendering.
Why this doesn't settle the AI question
It's tempting to read "Google removed it" as "it's dead, drop it." That conflates two different consumers of your markup. Google Search's rich result renderer is a presentation layer — it decided FAQ snippets cluttered results and were being gamed. That's a ranking-and-UX decision, not a statement about whether structured Q&A data is useful for machine comprehension.
Generative engines read pages differently than the classic ten-blue-links renderer does. Retrieval-augmented systems — whether that's Perplexity's live search layer, Bing/Copilot's index, or whatever ChatGPT's browsing tool pulls from — benefit from content that's already segmented into discrete, self-contained question-answer units. That's exactly what FAQPage schema forces you to produce: a clean array of {question, answer} objects with no ambiguity about where one answer ends and the next begins. Even without a Google rich result reading the tag, the underlying HTML structure (headings, short paragraphs, one idea per block) is the same thing GEO practitioners recommend for answer-block formatting generally.
The platforms themselves disagree publicly on how much this matters. Google's own guidance for AI Overviews and AI Mode states plainly that structured data isn't required for generative search and there's no special schema.org markup needed. Microsoft takes the opposite public position, describing schema as code that helps both search engines and AI systems understand content. Neither statement is independently verifiable from outside the companies, so treat both as vendor framing — Google has an incentive to simplify guidance and reduce spam-schema submissions, Microsoft has an incentive to position Bing/Copilot as more structured-data-friendly to attract technical SEOs. The honest answer is nobody outside these companies has clean experimental evidence either way.
When to keep FAQ schema on a page
Keep it where the markup is describing content a human would actually want to read as FAQ — pricing objections, setup troubleshooting, policy clarifications, "is X compatible with Y" comparisons. If you're already writing that content, wrapping it in FAQPage JSON-LD costs almost nothing and gives you a fallback structural signal in case any engine (present or future) does parse it preferentially.
Drop it where it was added purely to game the old rich-result snippet — thin, repetitive questions stuffed onto product pages that don't match how anyone actually talks. That content now has zero upside (no Google visual slot) and some downside (LLMs ingesting low-quality boilerplate under your brand's name doesn't help your citation profile).
How to write Q&As that match real prompts
The format matters less than the phrasing. Pull actual queries: look at your support tickets, sales call objections, and the "People also ask" boxes Google still shows outside the FAQ rich-result context. Write the question exactly as a user would phrase it conversationally — "does [product] work with Shopify" not "Shopify Compatibility" as a header. Answer in the first two sentences with the direct fact, then add detail after. This is the same pattern that works for AI Overviews extraction generally, independent of whether schema is wrapping it.
Avoid stacking ten generic questions at the bottom of every page template. Specificity per page beats a reused FAQ block — a reused block signals to both crawlers and readers that the content wasn't written for this page.
FAQ
Does removing FAQ rich results mean Google ignores FAQ schema for ranking?
Google has not said the markup itself is penalized — it withdrew the visual snippet because of overuse and intrusiveness in results, which is a presentation decision, not a content-quality signal.
Should I remove FAQPage JSON-LD from my site now?
Only remove it if the underlying Q&A content itself is low-value filler; if the questions and answers are genuinely useful, the schema costs nothing to keep and may still help AI retrieval systems parse the content cleanly.
Is FAQ schema required for AI Overviews or ChatGPT citations?
No — Google explicitly states structured data isn't required for AI Overviews or AI Mode, and there's no evidence any major AI engine requires FAQPage markup specifically to cite a page; well-structured plain content often performs just as well.
