THE SHORT ANSWER
The visual payoff disappeared in August 2023, when Google restricted FAQ rich results to a narrow set of government and health sites. The markup itself remains valid and cheap, and it gives machines an unambiguous question-to-answer pairing, which is a small retrieval convenience rather than a ranking lever. Add it if your platform emits it automatically. Do not build a project around it, and never mark up answers that a reader cannot see.
For about four years, FAQ markup was the closest thing SEO had to free real estate: a few blocks of JSON-LD and your result grew to twice the height of everyone else's. That era ended abruptly in August 2023, when Google narrowed eligibility to authoritative government and health sources and the expanded results vanished from commercial queries overnight.
What survived is a smaller and more honest question. The markup still validates, still parses, and still describes the structure of the page accurately. Whether describing that structure buys you anything now is the thing worth reasoning about rather than assuming in either direction, because the confident answers on both sides of this argument tend to come from people with something to sell.
The numbers, at a glance
What changed: Google restricted FAQ rich results in August 2023 to well-known government and health sites
What still works: the markup remains valid, parses cleanly, and pairs each question with its answer unambiguously
Realistic effect on citations: small and hard to isolate; retrieval reads your visible text regardless of whether it is annotated
Cost to maintain: near zero when generated from a content model, meaningful when hand-written per page
What the markup actually says, property by property
FAQPage declares that the page is a set of questions with answers supplied by the page owner. It is not for user-generated threads.
mainEntity holds the array of questions. Every question on the visible page belongs here; a subset invites an inconsistency flag.
Question with a name property carries the question text, which should be identical to the heading a human reads.
acceptedAnswer with an Answer node and its text property carries the answer. Limited inline HTML is permitted in that text field.
QAPage is the sibling type people reach for by mistake. It describes a single user-asked question with competing community answers, which is a forum, not a marketing page.
The rule that has not changed and never will: the marked-up text must be present and visible on the page. Annotating answers that only exist in the JSON-LD is a straightforward structured-data violation, and it is the one error in this area that carries an actual penalty rather than merely wasted effort.
Does it help you get quoted by an assistant?
Probably a little, and less than the people selling schema audits imply. Retrieval systems ingest rendered page text. A well-formed FAQ section with question headings and short answers is already highly extractable before any markup is applied, because the structure is legible in the prose itself.
Where the markup plausibly earns something is disambiguation at the parsing stage. An explicit acceptedAnswer node removes any question about which text belongs to which question, which matters on pages where the visual grouping is ambiguous or the layout is unusual. That is a real if modest benefit, and it argues for emitting the markup automatically rather than for hand-crafting it.
When to keep it, when to drop it
Keep it if your CMS or page generator produces it from the same source as the visible content, so drift is structurally impossible and the marginal cost is zero. Keep it on pages where the FAQ genuinely is the content rather than a bolted-on afterthought. Keep it if you operate in a health or public-sector context, where eligibility for the rich result may still apply.
Drop it if it is hand-maintained in a separate field that nobody updates when the copy changes, because a stale answer in the markup is worse than no markup. Drop it if the FAQ block exists only to host the schema, which is the pattern the 2023 change was aimed at and which still reads as filler to a human.
The thing worth doing instead
The durable version of this tactic is editorial, not technical. Real questions, phrased the way buyers actually phrase them, answered in forty to eighty self-contained words each, placed where a reader will find them. That structure is what gets lifted into an answer, and it works identically whether or not a single line of JSON-LD accompanies it.
For a contractor, the highest-value questions are almost always the awkward ones: what a job costs when it goes wrong, how long the wait really is, what is excluded from the quote. Those are the questions buyers ask an assistant precisely because they expect a supplier to dodge them, and answering them plainly is what makes a page the one worth quoting.
Deciding in fifteen minutes
Run three of your pages through a structured-data validator and note whether FAQPage is even present.
Compare each marked-up answer against the visible text and delete any that no longer match.
Check whether your platform emits the markup automatically or a human types it into a field.
Remove FAQ blocks that exist purely to carry schema and add nothing a reader wanted.
Rewrite the surviving answers to stand alone in forty to eighty words each.
Want leads like this in your pipeline?
Flock runs the campaigns, screens the enquiries and hands you only the ones that match your service area, job size and capacity. You pay per lead, not per month.
Book a 15-minute fit check | See lead package pricing
Related answers
Frequently asked questions
Will FAQ rich results ever come back?
Nobody outside Google knows, and the restriction has held for three years. Planning around a reversal is speculation. Build the FAQ so it earns its place through readability and extractability, then a reinstated rich result would be an unearned bonus rather than the business case.
Does FAQPage markup risk a penalty?
Only if the marked-up content is not visible on the page or the page is not genuinely a set of questions and answers. Both are documented violations. Correctly applied markup that no longer produces a rich result is simply inert, not harmful.
Should I use FAQPage or QAPage?
FAQPage when you wrote both the questions and the answers. QAPage when users ask and other users answer, as in a forum thread or a community board. Applying QAPage to marketing copy misdescribes the page and can trigger a structured-data warning.
Do AI assistants read JSON-LD at all?
Some pipelines parse it, others work purely from rendered text, and the mix is not publicly documented by any major vendor. Assume the visible text does the work and treat markup as a clarification layer rather than a delivery mechanism for content.
NEED A CLEARER PLAN?
Let’s turn your next move into momentum.
Talk to us →