What structured data AI engines actually read

Structured data is a machine-readable description of a page, usually written as JSON-LD, that states in plain terms what the page is about, and AI engines read it to extract your facts more reliably rather than to decide whether you get cited.

That distinction matters more than it used to, because the schema landscape changed twice in the last year and most published advice has not caught up. A lot of guides still tell you to add FAQ and HowTo markup as if it were 2022. Both have effectively been retired as visible search features. So before you spend an afternoon adding schema types, it helps to separate what engines still parse from what now earns nothing.

#The thing most guides get wrong

There are two separate jobs structured data does, and they are easy to conflate.

The first is rich results: the visual enhancements in a Google results page, like the stars under a product or the expandable questions under a result. The second is comprehension: giving any machine reading the page a clean, explicit statement of the entities and facts on it.

Google's own documentation is direct about the second job. It says Google "uses structured data that it finds on the web to understand the content of the page, as well as to gather information about the web and the world in general" (Google Search Central). That comprehension role does not disappear when a rich result does. A retired rich result means a SERP feature goes away. It does not mean crawlers stop reading the markup.

This is why the recent deprecations confuse people. The visible feature vanished, so the advice "add this schema" looks dead, but the underlying markup can still help a machine parse your page correctly.

#What now earns nothing as a rich result

Two waves of cuts reshaped the list.

In August 2023, Google restricted FAQ and HowTo rich results. FAQ enhancements were limited to "well-known, authoritative government and health websites," and HowTo was wound down across desktop and mobile (Google Search Central blog). For everyone else, those markups stopped producing the visual result they were added for.

Then in June 2025, Google deprecated seven more types: Book Actions, Course Info, Claim Review, Estimated Salary, Learning Video, Special Announcement, and Vehicle Listing. Google's stated reason was that these were not commonly used in Search and no longer added meaningful value (Search Engine Land).

The clearest signal came this year. FAQ rich results stopped appearing in Google Search on May 7, 2026, and Google confirmed sites do not need to rush to strip the markup out, noting that "unused structured data doesn't cause problems for Search" and that FAQPage remains a valid Schema.org type (Search Engine Journal).

Read that last point carefully. Google is telling you the feature is gone but the markup is harmless and still valid. That is the comprehension layer talking. The data can still describe your page to a machine even when no star or dropdown shows up.

#What still gets read

If you are optimizing for AI answers rather than chasing SERP decorations, a short list does most of the work. None of these depend on a deprecated rich result.

Article or BlogPosting with a real author and a dateModified. This is the markup that tells an engine who wrote a page and when it was last touched. For fast-moving topics, an undated page reads as "probably stale," and author attribution is the most legible trust signal you can hand a machine.

Product with explicit price, priceCurrency, and availability. When someone asks an assistant what a tool costs, an answer drawn from structured pricing is more accurate than one scraped from a rendered string. One caution: only use aggregateRating if you genuinely collect ratings. Google's guidance is explicit that you should not add structured data "about information that is not visible to the user" (Google Search Central), and fabricated reviews are a real penalty risk.

Organization with sameAs links. Pointing to your verified LinkedIn, GitHub, Crunchbase, and Wikipedia entry, where one exists, gives engines a graph that says "this entity is real and consistent across these surfaces." It is the closest thing to a machine-readable proof that you are a real company.

BreadcrumbList. Cheap to add and useful for letting an engine tell a section index apart from a citable leaf article.

The format question is settled. Google recommends JSON-LD "if your site's setup allows it, as it's the easiest solution for website owners to implement and maintain at scale" (Google Search Central). Use JSON-LD in a script block rather than inline microdata.

#The rule that keeps you out of trouble

Structured data must describe what a human can actually see on the page. Google states it plainly: "Don't create blank or empty pages just to hold structured data, and don't add structured data about information that is not visible to the user" (Google Search Central).

That single rule rules out most of the bad patterns: FAQ blocks stuffed onto marketing pages with no real questions, rating markup with no ratings behind it, prices in the schema that differ from the prices on screen. Engines have gotten better at catching the mismatch, and a mismatch hurts you more than missing markup does.

#How to validate

Two free tools cover it. The Schema.org validator checks syntax, and the Google Rich Results Test tells you whether a supported result would render. Google's full list of supported types is the canonical reference for what still produces a feature, so check it before you invest in any schema type a third-party guide recommends.

#The short version

Add Article and BlogPosting markup with author and dateModified to every content page. Add Product to pricing pages and Organization with sameAs once, site-wide. Keep BreadcrumbList if it is cheap. Treat FAQ, HowTo, and the seven types retired in 2025 as comprehension-only at best, not as features to chase. And never mark up anything a visitor cannot see. The schema list keeps shrinking, but the principle underneath it has not moved: give the machines reading your site a clean, honest statement of what is on the page.