FAQPage on the product page, or a separate Q&A page. Both lose..

On August 8, 2023, Google updated its search structured data documentation with a single decisive sentence. The sentence said that, from that point, `FAQPage` rich results would only be displayed for "well-known, authoritative government and health websites." For every other domain, the rich result disappeared. The schema validated. The stars and the dropdown that used to appear under search results did not.

Three years later, the schema is still in the wild on tens of thousands of ecommerce product pages, almost none of which qualify as authoritative health or government domains. The schema is shipped by SEO plugins, by review platforms, by themes, and by developers who pulled it from a tutorial written before the 2023 change. The schema produces zero rich results and almost no measurable benefit, while occasionally creating compliance liabilities under the FTC's testimonials rule when the "FAQ" questions are written by the brand and presented as if they were customer questions.

The companion schema, `QAPage`, has a different problem. `QAPage` was designed for sites like Stack Overflow and Quora, where one user posts a question and many users post answers. The rich result, when it appears, shows the question and the top-voted answer. Product pages, by their nature, do not have user-submitted questions with user-submitted answers; they have brand-curated content. Applying `QAPage` schema to a brand-curated FAQ is a different kind of policy violation, this time of the schema's stated semantics rather than of a Google rich result rule.

Most brands shipping question-and-answer content on product pages have chosen one of these two schemas. Both are wrong for the page they live on. The right answer is structural, not schema-level, and it depends on a question almost no operator is asked: does the brand have enough customer question volume to justify a separate page at all?

The `FAQPage` schema was added to Schema.org in 2018 and adopted into Google's rich results programme in May 2019. The intended use, per the original documentation, was "a page where the questions and the answers are written by the publisher." Wikipedia FAQs. Software documentation. Government health pages explaining a benefit programme. Pages where one canonical author had written one canonical answer to a frequently-asked question.

For its first four years, the schema worked. A site could mark up an FAQ section, get a dropdown rich result under its search listing, and gain roughly 15-25% more clicks on commercial intent queries, according to multiple case studies published in 2020 and 2021. The rich result was so successful that, almost immediately, every ecommerce theme added an "FAQ block" to its product page template. Every SEO plugin added an FAQ schema injector. Brands began writing five "frequently asked questions" per product. None of the questions had actually been asked; they were SEO content dressed as FAQ.

The schema was abused on a scale that surprised even Google's own teams. By 2022, internal Google data (referenced obliquely in a public talk by John Mueller in early 2023) showed that the median FAQPage block on the open web was four made-up questions written by a marketing team, of which two were promotional ("Why is X the best Y on the market?") and one was a thinly-disguised CTA ("Where can I buy X?"). The rich result, by then, had become a noise generator. Google's August 2023 change scoped the result back to authoritative domains and, in effect, killed the schema's value for everyone else.

The schema itself was not deprecated. It can still be shipped. It will still validate. It will produce, in 99% of cases, no visible rich result and no measurable impact on click-through rate. The bytes go on the page; the result does not appear.

The `QAPage` schema was designed, explicitly, for community Q&A sites. The schema.org definition references "a website like Stack Exchange, Quora, or AnswerBag." The structural assumption is that one user posts a `Question`, multiple users post `Answer` blocks, and the page records both the asker's identity, the answerer's identity, and the relative authority of the answer (an "accepted" or "top-voted" answer).

The schema makes assumptions a typical brand-curated product FAQ cannot satisfy. The `Question` is supposed to have an `author` who is not the page owner. The `AnswerCount` is supposed to be a real count of distinct user-submitted answers. The "accepted answer," if present, is supposed to be the answer the asker (or community moderation) marked as correct. None of these are true for a brand's curated FAQ section, which is written by the brand, for the brand's products, by a single author who is also the page owner.

Brands that have applied `QAPage` schema to a brand-curated FAQ are misrepresenting the semantics of the schema. Google does not, in practice, hand out manual penalties for this. The downside is structural: the schema is unlikely to earn a rich result, because the page does not match the underlying pattern; and any AI search engine reading the schema as a citation signal will, correctly, treat the content as brand-generated rather than community-generated, which is a weaker citation primitive than the brand thinks it is shipping. See first person dated signed for why community-authored content carries more citation weight than brand-authored.

The cases where `QAPage` is legitimately the right schema for a product context are narrow. A site that hosts genuine customer-submitted questions, with genuine customer-submitted answers, on a dedicated URL per question, qualifies. Most commerce sites do not host this. Amazon's "Questions & Answers" section is the canonical positive example: customers ask, customers (and sometimes the seller) answer, and each Q&A unit has its own structured data block. Outside Amazon and a handful of large marketplaces, the pattern is rare.

The real decision is not "FAQPage or QAPage." It is "should question-and-answer content live on the product page at all, or on a separate page that the questions justify."

Two patterns work. Most brands have not chosen between them; they have shipped a half-version of both.

The first pattern is question-and-answer content inline on the product page, with no schema attached at all. The content is treated as prose. The questions and answers appear under the product description, often above the review section, often paired with the photographs that buyers ask about. The questions are written by the brand, sometimes informed by customer support tickets, sometimes informed by review-content mining (see the long tail begins inside the review). The brand makes no schema claim about the questions being user-submitted. The content's job is to answer questions that real buyers actually ask, in the place where they are ready to ask them. The traffic is internal; the format is prose; the legal risk is low; the rich result is nothing, because no schema is shipped.

The second pattern is a dedicated Q&A page per product, where customers can actually submit questions, the brand or other customers can submit answers, and each question lives at its own URL. The schema is `QAPage`, applied truthfully. The rich result, where it appears, is real. The content is genuinely community-generated and therefore carries citation weight. The cost is operational: someone has to moderate, someone has to answer, and the page only earns its keep if there is real question volume.

The second pattern is the right answer only if the brand has the question volume to justify it. For a brand with one hundred SKUs and twelve customer questions a month, a dedicated Q&A page per product would mean ninety-eight empty pages and two pages with one question each. Empty pages are worse than no pages; they signal thinness to crawlers and create internal link bloat with nothing on the other end. For a brand with twelve SKUs and a hundred customer questions a month, the dedicated page becomes worth it. The threshold is not a hard number; it is the point at which the question volume per product page exceeds the operational cost of moderating it, which is usually somewhere between five and ten substantive questions per product per quarter.

Below the threshold, the right answer is inline prose with no schema. Above the threshold, the right answer is a dedicated page with `QAPage` schema, real question submission, and a moderation queue.

The reason most brands ship a wrong schema is that they confused the order of the decisions. The decision to add an FAQ section to the product page was made first, usually as a Conversion Rate Optimisation experiment, usually because a competitor had one. The decision to add schema was made second, usually by the same SEO plugin that shipped `Organization` schema with an `aggregateRating` (see self serving review markup and the line google drew). The brand never asked whether the FAQ content was real or whether the schema fit. The schema was a side-effect of a CRO decision.

Reversing the order produces a different answer. The first question is whether real customers are asking questions about the product, in volume that justifies an answer surface. If yes, the second question is whether the answer surface should be on the product page (low volume, prose, no schema) or on a separate page (high volume, structured submission, real `QAPage` schema). The schema follows from the architecture, not the other way around. Most brands are shipping schema as the leading decision and discovering, three years later, that the schema is doing nothing.

The schema is downstream of the architecture. The architecture is downstream of the volume. The volume is something almost no operator has measured.

The 2024 FTC rule on consumer reviews and testimonials, which took effect October 21, 2024, prohibits the use of fake or AI-generated reviews and the misrepresentation of reviews as coming from independent consumers when they do not. The rule has been read, mostly, as targeting fake reviews on product pages. There is a quieter implication for FAQ schema.

A typical brand-curated FAQ section presents the questions as if they had been asked. The phrasing pattern is consistent: "How long does the serum last?" "Is it safe for sensitive skin?" "Where is it manufactured?" The questions are written by a marketing team. They are not attributed to any specific asker. They are also not, in most cases, literally false; the questions are real questions the brand could plausibly have been asked.

The risk emerges when the FAQ is wrapped in `QAPage` schema, which is supposed to imply user-submitted questions and user-submitted answers. Marking up brand-curated content with `QAPage` schema is a representational claim that the content came from independent users when it did not. The FTC's December 2025 warning letters explicitly cited "misrepresenting the source of testimonial content" as a violation pattern. The schema, in this reading, is a technical extension of the same misrepresentation. The risk is small and largely theoretical, but it is a risk that did not exist before the rule, and brands shipping `QAPage` schema on curated FAQs should at minimum stop describing the questions as if they had been submitted by customers.

`FAQPage` schema, by contrast, makes no representational claim about user submission. Its definition explicitly says the questions and answers are written by the publisher. Shipping `FAQPage` schema on a brand-written FAQ is honest; it just does not produce a rich result, which is the separate problem covered above.

The objection, in 2026, is that even if the FAQ rich result is gone, AI search engines might still cite FAQ content because the schema makes the question-answer pattern legible. This is partially true and mostly overstated.

The Ahrefs March 2026 study of 1.4 million AI Overview prompts found that AI Overviews cite question-formatted content at roughly the same rate as prose-formatted content, when controlled for relevance and source authority. The schema wrapper had no measurable independent effect. What did have an effect: whether the content answered the specific question asked, whether the source was perceived as authoritative on the topic, and whether the answer was specific and dated rather than generic. (For the broader citation pattern, see the citation economy.)

The practical implication is that AI engines cite the answer, not the schema. A well-written prose paragraph on a product page that directly answers "how long does this serum last after opening" will be cited as readily as the same paragraph wrapped in `FAQPage` schema. The schema is not the citation primitive; the answer is. This is, on reflection, the same conclusion the human web reached about FAQ schema in 2021: the marketing value was always in the answer, not in the markup.

The most common mistake in commerce-schema work is to ship a wrapper around content that does not exist. The FAQ section that nobody asked. The Q&A page with no questions. The `aggregateRating` block on a homepage that is reviewing itself. Schema is supposed to be a description of something real. When the something real is missing, the schema describes the absence, and the engines (search engines, answer engines, AI crawlers) read the absence accurately. They cite the brand that has answered a real question, on a real page, in real words. They skip the brand that has shipped a wrapper.

The right move in 2026 is to find the real customer questions, answer them in real words, and let the schema be the last decision instead of the first. Brands that do this end up shipping less markup, more prose, and more pages worth indexing. (See the half life of a product page for the compounding case.) The brands that keep shipping the wrong schema are not penalised; they are just quietly absent from the citations they wrote the schema to win.