Check this page with an assistantOpens a chat asking it to summarise this article and name the evidence behind each claim.
Claude opens with the prompt on your clipboard: Anthropic does not support prefilled prompts on the web, and we would rather copy it than ship a button that drops it.
What Happened, With Dates
FAQ rich results stopped appearing in Google Search on 7 May 2026. Google said it would drop the FAQ search appearance, the rich result report, and support in the Rich Results Test in June 2026, with Search Console API support for the type ending in August. Rankings are not affected; this is a change in how results look.
It follows a pattern rather than arriving out of nowhere. HowTo rich results were deprecated on desktop in September 2023. In June 2025 Google retired seven further types at once, including Book Actions, Course Info, Claim Review, Estimated Salary, Learning Video, Special Announcement, and Vehicle Listing. The direction has been consistent for three years: the set of markup that produces something visible keeps narrowing.
The reading that fits the evidence is that Google withdraws rich result types that stop earning their space. FAQ markup in particular became a commodity deliverable, added to pages with invented questions purely to occupy more of the results page, which degrades the feature for searchers. A feature that is gamed until it is no longer useful tends to be removed, and that is roughly what appears to have happened here.
Do Not Rip Out Your Markup
Google's position is explicit: FAQ structured data can stay in place, the markup will not cause problems, and it simply will not produce visible results in Search. Unused structured data has never been a problem for Search. Removing it is a migration with risk and no benefit.
There are two further reasons to leave it. Google is not the only consumer of structured data, and a question-and-answer structure expressed in markup remains a legitimate description of a page that genuinely contains questions and answers. And the underlying content benefit was never the rich result: pages organised as direct answers to real questions are easier to extract from, which is the property that now matters for generated answers.
What should change is your posture toward new markup. If a vendor proposes adding FAQ schema across a site as an SEO deliverable, that proposal is now selling something that produces nothing visible. Adding questions to pages purely to carry markup was always the wrong version of this work, and the deprecation removes the last argument for it.
What Still Renders, and Why
The surviving types cluster into three groups: commerce, media, and identity. Each gives searchers something a plain blue link cannot, which appears to be the durable test for whether a type keeps its place.
| Type | Status | Worth implementing? |
|---|---|---|
| Product, Offer, merchant listing | Supported, and feeds shopping surfaces | Yes, highest priority for stores; see Merchant Center |
| Review and AggregateRating | Supported, with strict eligibility rules | Yes, where reviews are genuine and first-party |
| Organization and LocalBusiness | Supported; also used for entity resolution | Yes, for identity reasons more than visual ones |
| Article, Video, Recipe, Event, JobPosting | Supported where the content type applies | Yes, if you genuinely publish that thing |
| BreadcrumbList | Supported | Yes, cheap and improves how the result reads |
| FAQPage | No rich result since May 2026 | Keep existing; stop adding it for SEO reasons |
| HowTo | Deprecated since 2023 | No |
| The seven retired in June 2025 | Withdrawn | No |
The Organization row is the one whose importance has moved rather than shrunk. It never produced much of a visible result, and it is now the markup that helps a system be certain which company a page belongs to. That is a different job from decorating a search listing, and it is the job that matters more as answers get synthesised.
The Rules That Have Not Changed
Whatever the type, the constant is that markup must describe what the page visibly contains. Structured data is a labelling system, not an assertion layer, and every enforcement action in this area has been about the gap between the two.
- Mark up what is visible, only.
Reviews that appear on the page, prices the page shows, an author the page names. Markup describing content a visitor cannot see is the definition of the problem, and it is checked.
- Do not self-serve review markup.
Review snippets have specific eligibility rules and a long history of abuse. Ratings a business writes about itself are the clearest violation available and the easiest to detect.
- Validate against the current tool, not memory.
Support changes. A type that validated in 2024 may no longer be reported, and the Rich Results Test tells you what Google currently renders rather than what Schema.org permits.
- Expect eligibility, not entitlement.
Correct markup makes a page eligible for a rich result. It does not guarantee one, has never guaranteed one, and vendors promising specific appearances are promising something outside their control.
- Keep it in sync when the page changes.
Stale markup that contradicts the visible page is worse than no markup. Price changes are the usual culprit, which is why template-generated markup beats hand-written almost always.
What This Means for Budgets
Schema implementation has been a reliable line item on SEO proposals for a decade, frequently priced as though every type carries value. Three years of withdrawals make the honest version of that line much shorter, and considerably cheaper.
For most small businesses the whole job is now Organization or LocalBusiness on a stable page, BreadcrumbList if you have hierarchy, Article where you publish articles, and Product if you sell things. That is a day of work on a template, not a retainer, and the approach for the small-business case is in schema markup for small businesses.
Be especially wary of anything sold as AI-specific schema. No major assistant has committed to a special markup format, and Google's own documentation says no special machine-readable files are needed for its AI features. That claim has the same shape as the one examined in our review of llms.txt: an artefact that is easy to sell precisely because its absence of effect is impossible for the buyer to observe.
Questions People Ask About Structured Data
- Did Google remove FAQ rich results?
Yes. FAQ rich results stopped appearing in Google Search on 7 May 2026, and Google said it would drop the FAQ search appearance, the rich result report, and support in the Rich Results Test during June 2026. Teams pulling FAQ data through the Search Console API were given until August 2026 to adjust those calls.
- Should I remove FAQPage markup from my pages?
No. Google's guidance is that the markup can stay: it will not cause problems, it simply will not produce a visible result in Search. FAQPage remains a valid Schema.org type, unused structured data has never been a problem for Search, and other systems may still read it. Removing it is work with no upside.
- Does losing a rich result hurt rankings?
No. Google described the FAQ change as a search appearance change rather than an algorithmic one, and said rankings are not affected. What changes is how your result looks, which can affect click-through rate. That is a real commercial impact, but it is not a ranking penalty and there is nothing to recover from.
- Which structured data types still produce rich results?
The types with durable support are the ones tied to commerce, media, and identity: Product, Review and AggregateRating, Article, Recipe, Video, Organization, LocalBusiness, Event, JobPosting, and BreadcrumbList. The pattern across the retirements is that types offering little differentiated value to searchers get withdrawn while transactional and identity types persist.
- Is structured data worth adding if there is no rich result?
Sometimes, for narrower reasons than vendors suggest. Organization and LocalBusiness markup helps systems resolve who you are, which matters independently of visual results. Product data feeds shopping surfaces. Beyond that, markup that produces nothing visible and describes nothing a machine needs is effort with no demonstrated return, and it should be priced accordingly.
