FAQ markup no longer earns anything. Stop adding it.

FAQ blocks are still sold as an SEO add-on: a handful of questions at the foot of every post, wrapped in FAQPage markup, billed as a route to bigger search results. That route was narrowed to almost nobody in 2023 and closed outright in May 2026. Markup asks for a feature, and a request for a feature that no longer exists buys nothing except upkeep. The lesson reaches well beyond FAQs. Markup should be generated selectively, for the features a page can actually earn today.

The feature was narrowed in 2023 and switched off in 2026

The history is short and sits on Google's own pages. In August 2023 Google announced changes to HowTo and FAQ rich results. From then on, FAQ rich results would only be shown for well-known, authoritative government and health websites, and for everyone else they would no longer appear regularly. HowTo results were cut back to desktop that same month and fully deprecated in September.

The FAQ result hung on for its narrow audience for nearly three more years. Google's documentation changelog records a deprecation notice on 8 May 2026, saying the feature would no longer appear in Google Search starting 7 May. In June the FAQ documentation itself was removed. The address it used to live at now redirects to the changelog entry recording its removal.

So a blog that added FAQ markup any time after mid-2023 was, unless it happened to be a government health portal, asking for something it could not receive. Since May it has been asking for something nobody can receive.

Markup is a request, not a ranking signal

This is the part that the add-on pitch blurs. Structured data does not tell Google your page is good. It tells Google your page is eligible for a particular display, and Google decides whether to show it. We made the same point when we argued that structured data has to be generated: markup buys eligibility and legibility, not position.

Google's 2023 announcement is precise about the downside, and it is worth quoting fairly because it cuts against alarmism. You can drop the markup, but there is no need to proactively remove it, because structured data that is not being used does not cause problems for Search. Nobody is penalised for stale FAQ markup. The same sentence goes on to say that such markup also has no visible effects in Google Search. Harmless and useless, together. That is not a verdict on the questions themselves. It is a verdict on the tag.

The add-on outlived the feature because the receipt looks fine

FAQ blocks survived as a product because every visible step still works. The markup validates, the questions render, the invoice arrives. The only thing missing is the outcome, and the outcome was always the thing nobody checked.

Google has since put its own warning in writing. Its June 2026 guidance on third-party SEO tools and advice tells site owners to check outside advice against official documentation, notes that Google does not evaluate third-party services, and warns that using a tool does not guarantee ranking success. An FAQ upsell in late 2026 is exactly the claim that guidance asks you to test.

We should concede what is good here. A page can genuinely need a section of questions. A pricing page that answers the five things every prospect asks on the first call is doing real work for readers, with or without markup. The problem is the generic block appended to every post because the template has a slot for it. Those questions get written to fill the slot rather than answer anyone, which is padding with a schema wrapper, and we have argued before that word count is a symptom, not a goal.

Generated markup has to be selective

The fix follows from treating markup as a list of requests. A generator should emit a type only when two things hold: the feature exists in Google's current list of supported structured data features, and the page is plausibly eligible for it. A blog post earns Article and Breadcrumb. It does not earn Recipe because it mentions lunch, and it does not earn FAQPage because the brief said to add questions.

Selectivity is also what makes deprecation cheap. When the markup comes out of a template, retiring a dead type is one change that applies to the whole archive at once. Pasted into three hundred posts by hand, it stays forever, harmless, for every future audit to explain.

There is an honest limit to this. Google's list is not the only consumer of schema.org vocabulary, and other systems that parse your pages publish no equivalent list of what they use. Our rule is to emit what a named consumer documents using, and to stop adding the rest rather than speculate on readers we cannot see.

Every type in your markup should name the feature it asks for

The test is short enough to run on any template. For each type your pages emit, name the search feature it requests, and check whether that feature still exists and whether this page could get it. If you cannot name one, the markup is decoration. Decoration in a format nobody displays is upkeep without a return.

None of this is a trick in either direction. As with AI-written content, Google's systems are not hunting for the wrong kind of markup. They simply ignore what they no longer use. The cost of the FAQ block was never a penalty. It was the belief that it was doing something.