Which Structured Data Still Earns a Rich Result in 2026

Google removed FAQ rich results on 7 May 2026. HowTo had already gone. Both are still perfectly valid Schema.org types, both still validate, and neither will get you a single extra pixel in a search result.

That combination is why stale structured data survives so long: nothing breaks, no report turns red, and the markup keeps passing every test you might run on it. The only signal is that the SERP feature it was written for no longer exists.

Key takeaways#

• FAQ and HowTo are absent from Google's search gallery entirely. Not deprecated-but-supported. Gone.
• Google's Search Console API support for the FAQ report was removed in August 2026.
• Invalid markup is not the risk. Unused markup is: it costs bytes and maintenance and returns nothing.
• Article, Breadcrumb, Product, Review snippet, Event, Video, Organization and about eighteen others still earn results.
• Organization and Person matter more than their SERP appearance suggests, because entity resolution feeds AI answers.

What is still in Google's gallery?#

Twenty-five types, as of August 2026. That is the authoritative list — if a type is not in the search gallery, Google does not render a rich result for it, whatever Schema.org says the type means.

The full set: Article, Breadcrumb, Carousel, Course list, Dataset, Discussion forum, Education Q&A, Employer aggregate rating, Event, Image metadata, Job posting, Local business, Math solver, Movie, Organization, Product, Profile page, Q&A, Recipe, Review snippet, Software app, Speakable, Subscription and paywalled content, Vacation rental, Video.

Most sites need three or four of them. The ones that earn their keep on an ordinary content or product site:

Type	What it earns	Worth it for
Article	Headline, date and image treatment	Every blog post
BreadcrumbList	A path instead of a raw URL in the result	Every page below the root
Organization	Knowledge panel and entity association	Once, site-wide
Product	Price, availability, rating	Anything sold
Review snippet	Stars	Genuine reviews only
Video	Thumbnail, key moments	Pages whose main content is video

BreadcrumbList is the most reliably underrated of these. It is cheap, it is almost never wrong, and it changes what the user sees in the result line itself.

Why did FAQ rich results go away?#

Because the feature was overwhelmingly used to inflate listings rather than to answer questions. Google narrowed FAQ rich results to well-known health and government sites in 2023, and removed the appearance entirely in May 2026, along with the Search Console report and Rich Results Test support.

The deprecation ran in stages, which matters if you are trying to work out why a report vanished:

1. 2023 — eligibility narrowed to authoritative health and government sites. Most sites lost the feature here, quietly.
2. 7 May 2026 — the search appearance dropped for everyone.
3. June 2026 — the Search Console FAQ report and Rich Results Test support removed.
4. August 2026 — Search Console API support removed.

If a dashboard you built stopped returning FAQ data this month, stage 4 is the reason.

Should I remove my FAQPage markup?#

Not urgently, and not for a penalty that does not exist. Remove it when the markup has a cost you can name: bytes on a page you are optimising, a duplicated answer that has to be edited in two places, or a validation step in CI that is now checking something nobody consumes.

The distinction worth drawing is between the markup and the content. FAQ sections remain valuable, and for a reason that has nothing to do with rich results: a question-shaped heading with a short answer beneath it is the shape that answer engines and People Also Ask extract from. That works off the rendered HTML, not the JSON-LD.

So the honest position on FAQ content in 2026:

• Keep the on-page FAQ. Question as a heading, answer in 40–55 words directly under it.
• Stop generating FAQPage JSON-LD for new templates. It earns nothing.
• Leave existing FAQPage markup unless it is costing you something specific.
• Do not build reporting on it. The API is gone.

We reached the same conclusion on this site from the other direction. Tool pages here emit FAQPage and blog posts do not, and the obvious-looking fix was to add it to posts for consistency. Checking the gallery first turned that into the opposite decision: the posts got FAQ sections in the HTML, and no new JSON-LD.

What about HowTo?#

Same outcome, earlier. HowTo rich results were removed from desktop in 2023 and are absent from every surface now. The type validates, and produces nothing.

The replacement is not another schema type. It is writing the steps as a numbered list with a verb at the start of each item, which is what gets lifted into an answer regardless of markup. A HowTo block wrapping a page that has no visible steps was never going to work anyway; a page with genuine ordered steps does fine without the block.

What actually matters more than the SERP feature now#

Organization and Person, and not for the knowledge panel. Structured data has quietly acquired a second job: it is how a machine resolves which entity a page is about. That resolution feeds retrieval, and retrieval feeds the answers assistants generate.

A consistent Organization node — one canonical name, one url, a logo, the sameAs links to profiles you actually control — is a small file that makes your site unambiguous to a system trying to decide whether two mentions refer to the same thing. Unlike a rich result, you cannot see it working, which is exactly why it gets skipped.

Three practical rules:

1. One Organization node per site, referenced by @id from everywhere else, not repeated with slight variations per page.
2. sameAs only for profiles you own. It is an identity claim, not a link-building slot.
3. Keep names byte-identical across the node, the page title template and the logo's alt text. Variation is what forces a machine to guess.

Validating without a Rich Results Test for your type#

The Rich Results Test only reports on types that produce rich results, so a valid FAQPage now returns nothing there and reads as broken. Two tools, two different jobs:

• Schema Markup Validator (validator.schema.org) checks the markup against Schema.org itself. Use it for anything outside the gallery.
• Rich Results Test checks Google eligibility. Use it for the twenty-five types above, and read "no items detected" as "not a Google feature" rather than "invalid".

Search Console's own reports are the third view and the only one that reflects your live pages at scale — a type appearing there is a type Google is actually parsing on your site.

The schema markup generator produces the JSON-LD for the types that still earn something, and the SEO analyzer shows what a crawler receives from a live page, which is where a client-rendered JSON-LD block turns out to be invisible.

A short audit#

Worth half an hour on any site that has been accumulating markup for a few years:

1. List every @type you emit. Grep the templates rather than trusting memory.
2. Check each against the gallery. Anything not on it earns nothing.
3. For each orphan, ask what it costs. Bytes, duplicated content, a CI step. No cost, no urgency.
4. Confirm the survivors are server-rendered. JSON-LD injected by client JavaScript is frequently not seen.
5. Check Organization is one node, not many. This is the most common quiet defect.
6. Delete reporting built on removed types. It is returning nothing already.

Step 4 catches more real problems than the rest combined. Markup that a crawler never receives fails silently and validates perfectly in a browser.

Frequently asked questions#

Will removing FAQ schema hurt my rankings?#

No. It was never a ranking signal, only a search appearance, and the appearance no longer exists. Removing it changes nothing except the size of your page.

Does invalid or unused structured data get a site penalised?#

Unused markup does not, and Google has said so directly. Deliberately marking up content that is not visible on the page is a different matter and has always been against the guidelines.

Which structured data type should a blog post use?#

Article plus BreadcrumbList, both server-rendered. Article handles the headline, dates and image; BreadcrumbList changes the URL line in the result. Adding more types to a post rarely does anything.

Is JSON-LD better than microdata?#

For anything you are writing now, yes. Google supports all three formats, but JSON-LD sits in one block separate from the markup, so it can be generated, tested and changed without touching the HTML.

Do AI assistants read structured data?#

They read the page, and structured data helps them resolve entities and relationships within it rather than acting as a direct input. That is why Organization and Person are worth getting right even though their SERP payoff is invisible.

Related reading#

• llms.txt vs robots.txt — the other file everyone maintains for no measured effect
• Next.js Metadata API for SEO — emitting JSON-LD server-side in the App Router
• Core Web Vitals
• Generate the surviving types with the Schema Markup Generator, the head tags with the Meta Tag Generator, and audit a live URL with the SEO Analyzer.

References#

• Google: search gallery of structured data types
• Search Engine Journal: Google drops FAQ rich results
• Schema Markup Validator
• Google: Rich Results Test