FAQ rich results are gone from Google. Are you removing FAQPage markup, keeping it for other engines and AI, or ignoring it?

1 reply

• 

  adminAdmin

  Original poster 13 hours ago

  The short answer: I would leave accurate FAQPage markup where it already exists, stop adding it to new templates by default, and remove it wherever it does not match the visible content. Google has said the markup can stay and that unused structured data causes no problems, so removal is not urgent. The “keep it for AI” argument is weaker than it sounds, because the published evidence points to visible content doing the work, with the markup at best a secondary hint for some systems.

  The more useful decision is not about the JSON-LD at all. It is whether the FAQ content on those pages still earns its place now that it no longer buys extra space in Google’s results.

  What actually changed, and what it breaks in your reporting

  When	Change	Practical effect
  2023	FAQ rich results limited to well-known, authoritative government and health sites	Most commercial sites lost the feature then, not in 2026
  7 May 2026	FAQ rich results stopped appearing in Google Search	Remaining government and health sites may see CTR shifts from this date
  June 2026	FAQ search appearance filter, rich result report and Rich Results Test support removed	No FAQ-specific validation or reporting in Google’s tools
  August 2026	Search Console API support for FAQ rich results removed	Dashboards pulling the FAQ search appearance will break or report nothing

  Those dates come from Search Engine Journal’s coverage of the documentation change, which also notes that Google gave no explanation and did not connect the removal to its AI features. If you run a government or health site that still had FAQ results, add a Search Console annotation at 7 May 2026 so a CTR change on those pages is not misread as a ranking loss.

  Why removal is not urgent, and when it is

  Google’s documentation says “FAQ structured data can stay in place”, Google has previously said “unused structured data doesn’t cause problems for Search”, and FAQPage remains a valid Schema.org type. If your CMS or plugin generates the markup automatically from real on-page questions and answers, keeping it costs close to nothing.

  Removing it is not free either. It takes developer time, and on many platforms FAQPage sits in the same JSON-LD block as Product, Organization or BreadcrumbList markup, so a careless template edit can break markup that still produces Google features. That is my reasoning about implementation risk, not a documented Google warning, but it is why I would not schedule a sitewide removal project just because the rich result disappeared.

  Removal becomes the right call when the markup is a liability rather than neutral:

  • The markup contains questions or answers that are not visible on the page.
  • The answers are out of date (pricing, eligibility, returns, policy), so any system that does read them picks up wrong facts.
  • A plugin injects the same generic FAQ block across hundreds or thousands of URLs.
  • The FAQ block exists only because it once won a rich result, and the page would be better without it.

  The “keep it for other engines and AI” case, weighed by evidence

  This is where the discussion tends to run on assumption. Here is what the available evidence supports, ranked roughly by strength.

  Evidence	Type	What it supports	What it does not show
  Google’s generative AI guide (May 2026): structured data “isn’t required for generative AI search, and there’s no special schema.org markup you need to add”	Platform documentation	No FAQ-specific benefit to expect in AI Overviews or AI Mode	Anything about non-Google systems
  Fabrice Canel (Microsoft Bing) at SMX Munich, March 2025: schema markup helps Microsoft’s LLMs understand content	Platform representative statement	Structured data can feed Bing and Copilot’s understanding of a page	Anything FAQ-specific, or any measured effect on citations
  Ahrefs (May 2026): 1,885 pages that added JSON-LD between August 2025 and March 2026, against 4,000 control pages. Citations moved up 2.4% in AI Mode and 2.2% in ChatGPT, both statistically indistinguishable from zero, and down 4.6% in AI Overviews, a small but significant decline	Third-party dataset with a control group	Adding schema did not lift citations for pages that were already cited	FAQ in isolation (schema types were pooled), effects beyond 30 days, or pages with no prior citations
  searchVIU (late 2025): data placed only in JSON-LD was not extracted by ChatGPT, Claude, Perplexity, Gemini or AI Mode when fetching a test page	Small controlled test	Live page fetches appear to rely on visible content	Whether markup is used at indexing or training time

  The mechanism helps explain why these findings do not contradict each other. Setting training data aside, AI answers that cite web pages broadly get content in two ways: from a search index built earlier, or from a live fetch at answer time. In a live fetch, the searchVIU test suggests the system reads what a user would see. In an index-based path, structured data may be parsed during indexing, which is the scenario Canel describes for Microsoft. ChatGPT search, for example, uses OpenAI’s own crawler and, according to OpenAI’s help page, sometimes partners with third-party search providers, so its path is not a single pipeline you can reason about cleanly.

  The visible question-and-answer text is present in both paths. The markup is present in one, and of the platforms above, only Microsoft has said it uses it. So if FAQ content helps a page get cited, the content is the more likely cause. Anecdotes along the lines of “we added FAQ schema and started getting cited” rarely separate the markup from the content that was added with it.

  Important distinction:

  FAQPage markup was an eligibility ticket for a display feature. Google’s documentation framed it as eligibility for a rich result, not as a ranking or citation factor. Losing the display changes how your listing looks, not what Google understands about the page.

  The better question: does the FAQ content still earn its place?

  A lot of FAQ blocks were written between 2019 and 2023 to capture extra SERP space. They tend to share the same pattern: generic questions, answers that restate the main copy, and identical blocks repeated across templates. With the rich result gone, those blocks have no display benefit left, and they still add length and repetition to the page.

  Good FAQ content looks different. It answers questions real customers ask (from support tickets, site search, sales calls or live chat logs), the answer is specific to that page, and it removes friction before a conversion: delivery times, returns, compatibility, eligibility, pricing conditions. That content is worth keeping for users regardless of markup, and it is also the kind of passage an AI system could plausibly quote.

  Hypothetical example:

  A SaaS company has 300 feature pages, each carrying the same six FAQs about trials and billing, plus a help centre with 150 distinct question-and-answer articles. The feature-page FAQs are duplicated boilerplate: I would consolidate them into one billing FAQ page and remove the blocks (and their markup) from the feature pages. The help centre articles answer specific questions: I would keep both the content and the markup, because it is accurate, costs nothing to maintain, and may help systems that do parse it.

  A decision framework by situation

  Situation	What I would do	Evidence to inspect first
  Markup generated automatically from genuine, visible, page-specific Q&A	Keep; no action needed	Spot-check that markup and visible text match
  Markup contains content not shown on the page, or stale answers	Fix or remove now	Crawl comparing JSON-LD answers to rendered text
  Same FAQ block repeated across many templates	Decide on the content first; the markup follows that decision	Duplication across URLs, engagement with the block, conversion paths
  New templates or new pages	Add FAQ content only where users need it; add markup only if it is automatic and accurate	Support and search data showing real questions
  Bing or Copilot is a meaningful channel for you	Keep accurate markup; Microsoft is on record saying schema helps its LLMs	Bing Webmaster Tools AI Performance citations

  If you want an answer for your own site, test removal

  Because the markup is already live, a removal test is cheaper and more reversible than an addition test, and Ahrefs has already covered the addition side. A workable design:

  • Hypothesis: removing FAQPage markup, with the visible FAQ content unchanged, does not reduce AI citations or organic performance.
  • Test and control: remove markup on one set of comparable pages or templates; leave a matched set untouched.
  • Primary metrics: page-level citations in the Bing Webmaster Tools AI Performance report (public preview, citations only, no clicks), plus Google’s Search generative AI performance reports (impressions only; available to all eligible sites worldwide since 31 August 2026).
  • Secondary metrics: organic clicks and conversions on both groups, and third-party AI citation tracking if you already pay for it, accepting that it is sample-based.
  • Duration: confirm recrawl of the test pages first, then run for at least six to eight weeks. Note any core update or major content edits during the window.

  What would change the decision: if the test group loses citations relative to control by more than the normal week-to-week movement in the control group, keep the markup and treat it as useful for your mix of engines. If there is no difference, the markup is neutral for you and the choice comes down to maintenance cost. On a smaller site, citation counts may be too sparse to detect anything, in which case I would not bother testing and would simply keep accurate, automated markup.

  My position

  I am not removing FAQPage markup for its own sake, and I am not keeping it on the belief that it drives AI citations, because nothing published so far shows that. I am auditing the FAQ content: keeping blocks that answer real questions (with accurate markup attached), cutting boilerplate blocks that only existed for the rich result, and fixing any markup that says something the page does not. That turns a markup question with little evidence behind either side into a content decision you can measure through engagement, conversions and, where you have the data, citations.

  Need help with this on your own site?

  Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.

  See SEO services