FAQ and HowTo Schema After the Deprecations: What Still Earns Visibility
In August 2023 Google did something it rarely does: it took rich results away. FAQ rich results — the expandable question-and-answer accordions that had spread across half the web’s content pages — were restricted to well-known, authoritative government and health sites. HowTo rich results were deprecated outright shortly after, and then removed across desktop and mobile. Overnight, the structured data that millions of pages had added specifically to earn those snippets stopped producing them. The reflexive response was to rip the markup out. That was the wrong response, and understanding why requires separating what the deprecation actually removed from what the markup still does.
What Google Actually Cut
The precise scope matters because the panic overshot it. Google did not declare FAQPage or HowTo invalid, harmful, or penalized. It stopped rendering the rich result for them on the general web. FAQ rich results were narrowed to authoritative government and health sites; for everyone else, the expandable Q&A accordion no longer appears in the search result. HowTo rich results, which had shown stepped instructions with images directly in the SERP, were deprecated and then dropped entirely.
The markup itself is still valid schema.org vocabulary, still parsed, and still understood. Google did not say “remove this”; it said “we will no longer show a special result for it.” That is a narrower statement than the way it was received. A page that carries accurate FAQPage markup today is not being penalized for it and is not violating anything. It simply will not get the SERP accordion it once did — and if the markup was only ever there to earn that accordion, its job is done and removing it is reasonable. The error is concluding that the deprecation made the markup worthless. It made one specific output of the markup unavailable, while leaving the markup’s other functions intact.
The reason Google gave is also instructive: the FAQ and HowTo results had become cluttered and over-used, often added to pages where the Q&A or step structure was a markup convenience rather than the page’s genuine purpose. The deprecation was a cleanup of an over-gamed feature, not a verdict that structured Q&A content has no value. That distinction is the whole basis for what still works.
Where the Markup Still Earns Visibility
Strip away the Google rich result and three real functions remain, and on the current trajectory they are growing more important, not less.
AI answer engines consume structured Q&A. An answer engine grounding a response benefits from content already decomposed into explicit question-and-answer units, because that structure maps directly onto how the engine retrieves and presents answers. FAQPage markup labels exactly which text is a question and which is its answer, in a machine-readable form, removing the inference the engine would otherwise have to make from prose. The Google SERP accordion is gone; the underlying signal — “this is a discrete question and this is its self-contained answer” — is exactly the signal AI extraction rewards. The markup outlived the rich result it was built for because a different class of consumer started using the same structure.
This is also why the answer-block discipline and FAQ markup are complements, not substitutes. The retrievable answer has to exist in the visible body content; the markup then labels it explicitly. The markup does not rescue a buried or JavaScript-injected answer — the answer must be present in the raw HTML for any crawler that doesn’t render — but where a real, self-contained answer exists in the text, FAQPage markup makes its structure unambiguous to the systems reading it.
Non-Google engines and surfaces still use it. Google’s deprecation governs Google. Other search engines, vertical search tools, and downstream consumers of structured data did not all follow, and several continue to use FAQPage and HowTo to render or organize Q&A and step content. Removing valid markup that accurately describes your content sacrifices whatever value those other surfaces still extract from it, in exchange for nothing.
It documents content structure cleanly. Even setting aside any consuming engine, accurate FAQPage and HowTo markup is a faithful machine-readable description of the page’s structure. That is the original purpose of structured data — describe the page so machines understand it — and it remains valid whether or not any particular engine currently renders a special result. The markup’s job was never solely to earn an accordion.
The decision rule that follows: keep FAQPage and HowTo markup where it accurately describes genuine Q&A or step content on the page, treating it as an extraction and description aid rather than a rich-result lever. Remove it where it was bolted onto pages whose purpose isn’t really Q&A, or where it was added purely to game the now-gone result. Accuracy is the test, not the rich result.
QAPage and DiscussionForumPosting: The Types That Kept Their Results
The deprecation did not touch every Q&A-shaped type, and the survivors point to where Google actually wants this structure used. QAPage and DiscussionForumPosting retained rich-result support precisely when FAQPage and HowTo lost theirs, and the reason is the content they describe.
FAQPage was always intended for official, site-authored Q&A — the questions a publisher writes and answers about its own product or topic. That is the category Google decided was over-used and cluttered. QAPage, by contrast, describes a page built around a single user-posed question with community answers — a forum thread, a Stack Overflow-style page, a product Q&A section where real users ask and answer. DiscussionForumPosting describes forum-style posts and their replies, the structure of genuine community discussion. These are user-generated content, and Google’s interest in surfacing authentic community discussion grew over the same period it was deprecating site-authored FAQ accordions — the well-documented rise of “people want forum answers” behavior.
The practical guidance is to match the type to the content’s true nature. If real users are asking and answering questions on the page, that is QAPage or DiscussionForumPosting, and those types remain eligible for rich treatment. If your editorial team wrote the questions and answers, that is FAQPage, which no longer earns the general-web rich result and should be kept only as an extraction and description aid. Using QAPage to dress up site-authored FAQ content as community Q&A is a misrepresentation of the content type — the same kind of mismatch the deprecation was meant to clean up — and risks the markup being ignored or worse. The type must describe what the content actually is.
digitalSourceType for AI-Generated Content Disclosure
The newer piece of this vocabulary addresses a problem the 2023 deprecation predates: how to disclose, in structured data, that content was produced or assisted by AI. digitalSourceType — drawn from the IPTC media provenance vocabulary and usable in schema.org contexts — labels the production method of content: created by a human, generated by a machine, a composite, an algorithmically enhanced version, and so on.
The relevance to this discussion is provenance honesty. As AI-generated content proliferates, and as both search and answer engines weight first-hand human evidence more heavily, disclosing the production method of content is a trust signal rather than a confession. A digitalSourceType value declaring that an image or passage was AI-generated, or human-created, makes the provenance machine-legible instead of leaving it to detection heuristics. It is a structured-data complement to the author and E-E-A-T signals that establish who stands behind a page: where those say who produced the content, digitalSourceType can say how it was produced. Treat it as an emerging disclosure mechanism — adoption and consumption are still maturing — but one aligned with the direction provenance is heading, where labeled honesty about AI involvement is an asset, not a liability.
The Net Position
The 2023 deprecations removed two SERP features and changed nothing about the validity or usefulness of the underlying markup for its other functions. FAQPage and HowTo no longer earn the general-web rich results they were famous for, so do not add them expecting that accordion — and clean them out where they were only ever there to game it. But where they accurately describe real Q&A or step content, keep them as aids to AI extraction, non-Google surfaces, and machine-readable description. For genuine user-generated Q&A, use QAPage and DiscussionForumPosting, which kept their rich-result eligibility and match the community content Google actually wants to surface. And as AI-produced content grows, digitalSourceType offers a structured way to disclose provenance honestly.
A visibility platform helps by auditing structured data against this post-deprecation reality: flagging FAQPage and HowTo markup added in pursuit of rich results that no longer render, confirming the markup that remains accurately describes genuine Q&A or step content rather than misrepresenting the page, checking that user-generated Q&A uses QAPage or DiscussionForumPosting rather than FAQPage, and verifying that the answers the markup labels actually exist in the raw HTML an AI crawler receives. It cannot promise that any engine will extract or cite a given answer — that decision belongs to the engines — but it can keep your schema honest, current, and matched to the content it describes, which is what keeps it useful as the surfaces that consume it shift.