Structured Data in 2026: What Still Earns Rich Results (and What Google Quietly Retired)

For most of the last decade, structured data was treated as a checkbox: install a plugin, let it stamp FAQPage and HowTo markup onto every page it could reach, and collect the extra SERP real estate. That trade stopped paying. On 7 May 2026, Google added a deprecation notice to its FAQ structured data developer documentation, and FAQ rich results no longer appear in search results at all. HowTo went the same way — as of May 2026 there is no HowTo rich result on any Google surface. The markup is still valid schema.org. It just does nothing.

This matters more than a lost star rating, because the plugins that generated all that markup are still generating it, and the reports people used to measure it are being switched off on a schedule. The teams that notice will spend the second half of 2026 rebuilding their schema around the types that still carry weight. The teams that don't will keep shipping bytes that no search engine acts on, and keep reading dashboards that no longer report anything.

What actually changed, and on what timeline

The retirement is being staged rather than flipped off in one go, which is part of why it slipped past so many people. The deprecation notice landed in the documentation in May 2026. In June 2026, the FAQ search appearance filter, the FAQ rich result report in Search Console, and Rich Results Test support for FAQPage were all removed. The FAQ rich result data in the Search Console API follows in August 2026. So a site can still be looking at historical FAQ impressions in a report that is, by the time anyone reads it, describing a search feature that no longer renders.

The broader context is Google's March 2026 core update, which reduced rich result display for several schema types that had been widely abused — FAQ, Review, and HowTo bolted onto pages whose primary content had nothing to do with them. Sites whose schema described what the page genuinely was kept their rich results, and in a number of cases improved. Sites that used schema as a growth hack lost the surface entirely. That's the through-line: Google didn't stop reading structured data, it stopped rewarding structured data that misrepresents the page.

What still works

The types that continue to generate rich results and click-through gains are the ones that describe a page's actual substance rather than annotate it: Article, Product, Review (as first-party review content, on the thing being reviewed), BreadcrumbList, Video, and Event. Organization and LocalBusiness remain the backbone of entity understanding — they're how a search engine knows which company a domain belongs to, how to reconcile your name across the web, and what to show in a knowledge panel. None of these are new or exotic. They're the boring half of the schema vocabulary that never needed a plugin to justify itself.

There's a second reason to care about them in 2026 that didn't exist in 2020. Structured data is one of the clearest signals available to the systems assembling AI Overviews and generative answers. When a model needs to establish what a page is, who published it, when it was updated, and what entity it concerns, explicit markup answers those questions unambiguously instead of leaving them to inference. We wrote about generative-search visibility earlier this year, and schema is the least glamorous and most reliable lever in that whole conversation.

Why plugin-generated schema is the problem, not the solution

Here's where this stops being an SEO housekeeping story and starts being an architecture one. On a page-builder or template site, structured data is produced by a plugin inferring intent from markup that was never organized with that intent in mind. It sees an accordion and emits FAQPage. It sees a numbered list and emits HowTo. It emitted both of those enthusiastically for years, which is precisely why Google is now discounting them. The plugin cannot distinguish a page that genuinely is a product page from one that merely contains a price, because it has no model of the site — only of the DOM.

A custom build inverts that. In a Next.js application, the data that drives a page already exists as typed content before anything renders: a product has a price, a currency, an availability state; an article has an author, a publish date, a canonical URL. Emitting JSON-LD from that source of truth is a few lines of server-rendered code, and it is correct by construction — it cannot drift from the visible content, because it's generated from the same object. When Google retires a type, you delete one function. When a new one becomes worth emitting, you add one. Nobody audits a plugin's output page by page hoping it guessed right.

The audit worth running this month

Three things are worth checking before Q4. First, inventory what you're actually emitting: view source on your key templates, or run a handful of URLs through Google's Rich Results Test, and see whether FAQPage and HowTo are still on the page. Leaving them costs nothing directly — they're valid schema.org and won't trigger a penalty — but their presence usually indicates a plugin operating on assumptions nobody has revisited, and that same plugin is likely making other calls you'd disagree with.

Second, check whether the types that still matter are present and complete. A partially populated Product or Article block is common and quietly ineffective; missing a required property means no rich result even though the markup validates as schema.org. Third, reconcile your reporting. If a chunk of the traffic in your SEO dashboard was attributed to FAQ rich results, that line is going to zero for reasons that have nothing to do with your content quality, and you want to know that before someone reads the drop as a ranking problem and starts changing things that were working.

What this pattern keeps telling us

This is the third time in about eighteen months that the same shape of change has hit the same kind of site. Core Web Vitals and INP rewarded builds where performance was a rendering decision rather than a plugin. Accessibility enforcement rewarded builds where semantic markup was written deliberately instead of patched at runtime. Structured data is now rewarding builds where schema is generated from real content models instead of guessed from layout. In every case, the requirement moved from something you could bolt on to something that has to be true of the code.

That's not a coincidence, and it's not really about search engines. It's what happens when the systems consuming your site get good enough to tell the difference between a page that is something and a page that is decorated to look like it. If your structured data is currently produced by a plugin you haven't opened in two years, the May 2026 deprecation is a cheap warning. The next one won't necessarily be.