A technical reference on AI search visibility. This site sells nothing, takes no engagements and endorses no products. Consulting enquiries are handled separately at hartzer.com.

Hartzer.it.com logoHartzer.it.comAI search visibility reference
Abstract stepped block illustration representing FAQ Markup and Answer Extraction
Technique or Signal

FAQ Markup and Answer Extraction

The 2023 carve-out for government and health sites is gone. Most FAQ advice published in 2026 is describing a feature that no longer exists.

UnsupportedFAQ rich results were removed from Google Search entirely on 7 May 2026; the markup now produces no search appearance at all.

Three things get called FAQ optimization

They have different mechanisms and, since May 2026, very different fates. Keeping them apart is most of the work on this subject.

FAQPage structured data is schema.org markup declaring that a page contains a list of questions and their answers, each question a Question with an acceptedAnswer. Between 2018 and 2023 it made a page eligible for an FAQ rich result — an expandable accordion attached to the page's listing in Google's results, which visibly enlarged the listing and was widely credited with lifting click-through.

Question headings are an editorial pattern, not markup: writing an h2 or h3 as the question a reader would actually type, and answering it in the sentence or two immediately beneath. No schema is involved and none ever was.

Answer extraction is what a search or AI system does to your page — pulling a span of text out of it and presenting that span as the answer, with or without a link back. It is an applied form of question answering, the long-standing information retrieval problem of returning an answer rather than a list of documents. Featured snippets, People Also Ask, and the sentence-level quoting inside an AI Overview are all answer extraction.

As of August 2026 the markup is dead as a Google search feature, the editorial pattern is not, and they were never the same thing. Almost every expensive mistake on this subject comes from treating the first as the cause of the third.

7 May 2026: the removal that arrived without an announcement

FAQ rich results stopped appearing in Google Search on 7 May 2026. A deprecation notice was added to Google's FAQ structured data documentation the following day, reading in part:

FAQ rich results are no longer appearing in Google Search. We will be dropping the FAQ search appearance, rich result report, and support in the Rich results test in June 2026. To allow time for adjusting your API calls, support for the FAQ rich result in the Search Console API will be removed in August 2026.

Google followed the schedule. The documentation itself was removed on 15 June 2026, FAQ vanished from the structured data gallery in that day's refresh, and Search Console API support ended in August 2026. What remains is 25 structured data features in Google's gallery, with neither FAQ nor HowTo among them.

There was no blog post. Matt G. Southern reported the change for Search Engine Journal on 10 May 2026 and noted that Google did not publish a blog post or explain the reasons behind the removal. The only announcement was a notice inside a documentation page, and that page was itself deleted five weeks later. A documentation notice is an in-place edit to a reference document rather than a broadcast, and it reaches only the people who happened to reload that URL in a five-week window.

That is the mechanical explanation for why this change has propagated so slowly through practitioner knowledge, and why so much advice written after it is simply wrong.

Why 2026 advice still repeats the 2023 carve-out

The August 2023 change was a reduction, not a removal. Google announced on 8 August 2023 that FAQ rich results would be shown only for well-known, authoritative government and health websites, citing a cleaner and more consistent search experience, and restricted HowTo to desktop before removing it entirely on 14 September 2023.

That carve-out became the standard sentence in every FAQ article written between 2023 and 2025: FAQ rich results are now limited to government and health sites. It was accurate for two years and nine months. It has been wrong since 7 May 2026. There is no carve-out. There is no government-and-health exception. The feature does not exist for anyone.

Alongside it, Google's 2023 guidance on what to do with markup that no longer produces anything — that while you can drop this structured data there is no need to proactively remove it, and that structured data which is not being used does not cause problems for Search — is why millions of FAQPage blocks are still deployed today. The advice was correct. It meant harmless. It was widely read as still working, and that misreading is now two years and one full deprecation old.

Anyone whose knowledge of this topic stops at 2023 is describing a feature with a carve-out that has been gone for over a year, and the dated documentation changelog is the primary record of both steps.

Does FAQPage markup help an AI system cite you?

No, and three independent lines of evidence point the same way.

Google states there is no special schema.org structured data you need to add for its AI features, on a page stamped 10 December 2025, and repeats it in the May 2026 generative-AI guide under a section it titled Mythbusting generative AI search: what you don't need to do. Ahrefs' controlled test of 1,885 pages adding JSON-LD against roughly 4,000 matched controls, published 11 May 2026, found citation changes of −4.6% on AI Overviews, +2.4% on AI Mode and +2.2% on ChatGPT. And no AI operator documents parsing it: OpenAI, Anthropic and Perplexity all publish crawler documentation and none mentions FAQ markup, question headings or Q&A structuring, while Bing's guidelines, rewritten in February 2026 to cover Copilot, make no such claim either.

The counter-claim in circulation is a 112% lift in AI citations from FAQ schema. It comes from an agency study by TechCognate on how Gemini selects citations, based on roughly a thousand pages. It is a correlation between pages that get cited and pages that carry FAQ markup, and in most downstream repetitions the qualifier is gone. It is contradicted by the only controlled test on the question. What would settle it definitively: a matched test adding only FAQPage markup with the visible text held constant. Nobody has run one.

There is one more piece of negative evidence, and it is the quietest. Google's May 2026 guide was written specifically to tell publishers what does and does not help with generative AI features. It does not mention FAQ, question format or Q&A anywhere, in either direction. Google had the opportunity to say that question-and-answer formatting helps and declined to take it.

The mechanism that survived

Featured snippets and People Also Ask select a passage from a page's visible text and present it as the answer. Google's documentation has been consistent on one point for a decade: there is no markup that triggers one. Selection is algorithmic. The only publisher controls are suppressive — nosnippet, data-nosnippet and max-snippet. The same holds one layer up: the sentences an AI Overview quotes are drawn from retrievable text, not from a JSON-LD block.

Which leaves the editorial pattern, and it has a real argument behind it that has nothing to do with schema. A question heading followed by a direct answer produces a self-contained span of text — one that means the same thing out of context as in it. Retrieval systems operate on spans. A paragraph opening it depends on several factors, as we discussed above is worthless when retrieved alone; a paragraph opening with the answer is not.

That argument is about the retrievability of text and it holds regardless of what happened to the markup. It is also close to Google's own framing, which endorses organizing pages into paragraphs, sections and headings that give content a clear structure — while grounding that endorsement in what people appreciate rather than in what machines parse. Google goes further in the other direction in the same document:

There's no requirement to break your content into tiny pieces for AI to better understand it.

Read the two together and the instruction is clear. Structure a page so a reader finds the answer. Do not shred it into micro-answers because you believe a machine wants that. Label the pattern honestly: it is a sound editorial default with a plausible retrieval mechanism, not a documented ranking factor. No operator has published data attributing citations to question headings.

What to do with the FAQ blocks you already have

The practical decisions are short, and most of them are decisions not to act.

  • Leave existing FAQPage markup in place. Google's standing position is that unused structured data does not cause problems for Search, and the coverage of the 2026 change reiterated that sites do not need to rush to remove it. FAQPage remains a valid, non-deprecated schema.org type at v30.0, released 19 March 2026, with reported adoption across one to ten million domains. Removal is optional cleanup, not remediation, and it costs template work and regression risk for a benefit of zero.
  • Stop deploying new FAQPage markup as a Google Search deliverable. There is no surface for it to produce. The feature, the report, the validator support and the documentation are all gone. Billing for FAQ schema implementation as a search-visibility improvement in 2026 is billing for something that cannot work.
  • Do not substitute QAPage. It is a different type, documented for pages where a user posts a question and a community answers — forums and community sites — and Google still lists Q&A in its search gallery. Applying it to editor-written marketing copy misrepresents the page. Structured data must match the page's actual nature, and that is a spam-policy line rather than a technicality.
  • Keep writing the questions readers actually ask. A question heading with the answer directly beneath is good writing and produces retrievable passages. That is the half of FAQ practice that never depended on Google.
  • Do not bolt a question block onto the bottom of a page. Padding with questions nobody asked, answered in terms already covered above, adds text that is redundant with the page's own body. It dilutes the page rather than sharpening it.

The pattern in this whole episode is worth naming, because it will repeat. A feature is announced loudly, restricted quietly, removed silently, and its documentation deleted — while an industry keeps selling implementation of it for three more years. The one piece of contemporaneous reporting and a dated changelog entry are the entire public record of a change that invalidated a common line item on thousands of invoices.

Frequently asked questions

Is FAQ schema still worth adding in 2026?

Not for Google Search. FAQ rich results stopped appearing on 7 May 2026, the documentation was removed on 15 June 2026, and Search Console API support ended in August 2026. There is no surface left for the markup to produce. It remains a valid schema.org type, so a non-Google consumer could theoretically parse it, but no one has announced that they do.

Are FAQ rich results still shown for government and health sites?

No. That carve-out came from Google's 8 August 2023 announcement and it ended on 7 May 2026, when the feature was removed from Google Search entirely. Advice published in 2026 that still repeats the government-and-health exception is describing a state of affairs that no longer exists. There is no site category that still receives FAQ rich results.

Should I remove FAQPage markup from my pages?

There is no need to. Google's position is that structured data which is not being used does not cause problems for Search, and contemporaneous coverage of the 2026 removal reiterated that sites do not need to rush to strip it. Removal means template changes and regression risk for no benefit. Leave it; just stop treating it as a deliverable.

Does FAQ markup increase citations in AI Overviews or ChatGPT?

No published evidence supports it. Google says no special schema.org markup is required for its AI features, the one controlled test of adding JSON-LD found no uplift on any platform, and no AI operator documents parsing FAQ markup. The widely quoted 112% figure is correlational, comes from an unattributed agency study, and is contradicted by that controlled test.

Can I mark up a passage so Google uses it as the answer?

You cannot. Google's featured snippet documentation has consistently said that its systems decide whether a page would make a good snippet and elevate it algorithmically. There is no markup that forces extraction. The only publisher controls are suppressive — nosnippet, data-nosnippet and max-snippet — and they remove you from the feature rather than qualify you for it.

Do question headings help content get quoted by AI systems?

Consistently reported by practitioners, with a sound mechanism and no platform confirmation. A question heading answered directly beneath produces a self-contained passage that survives being retrieved alone. Google endorses paragraphs, sections and headings, but frames it as something people appreciate, and adds in the same guide that there is no requirement to break content into tiny pieces.

Top