Structured data for AI search: which schema still earns its place

What structured data can and cannot do

Search engines and AI systems read your visible text first. Structured data is a second, machine-readable statement of the same facts. It helps in three ways:

Disambiguation. A name like "Mercury" could refer to many things. An Organization or Person block with sameAs links to profiles such as LinkedIn, GitHub or Wikidata ties the name to one entity.
Precise facts. Prices, availability, dates, ratings and authorship are easy to misread in prose and unambiguous in markup.
Rich results in classic search, which can still earn clicks alongside AI answers.

Google's guidance for its AI features says structured data isn't required and that you should make sure any structured data matches the visible text on the page. Microsoft has gone further: at SMX Munich in March 2025, Bing's Fabrice Canel said schema markup helps Microsoft's language models understand content. Neither company says markup earns citations on its own, and it cannot make thin content worth citing. Markup that contradicts the visible content can be ignored, and Google treats misleading markup as spam.

The types worth maintaining

Type	Use it for	Fields that matter most
Organization	Your company or publication, once per site (usually on the home page)	name, url, logo, sameAs, contactPoint
Person	Authors and named experts	name, jobTitle, url, image, sameAs, worksFor
Article / BlogPosting	Guides, posts and news	headline, author (linked to the Person), datePublished, dateModified, image
Product + Offer	Anything with a price	name, brand, offers.price, offers.priceCurrency, offers.availability
BreadcrumbList	Every page below the home page	the path from home to the current page
Book	Books and long-form guides	name, author, bookFormat, numberOfPages, isAccessibleForFree, url
SoftwareApplication	Apps and tools	name, operatingSystem, applicationCategory, offers

Two former favourites have lost their search features. Google withdrew HowTo rich results in September 2023, and it stopped showing FAQ rich results for every site from 7 May 2026. Both remain valid schema.org vocabulary, and FAQPage still describes question-and-answer content cleanly, so keep it where a page really is a list of questions and answers. Just do not expect a visual result in Google for it.

Connect the entities

Separate blocks are more useful when they point to each other. Give your main entities stable identifiers with @id and refer to them:

{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Person",
      "@id": "https://www.example.com/#author",
      "name": "Jane Example",
      "jobTitle": "Head of Search",
      "sameAs": ["https://www.linkedin.com/in/jane-example/"]
    },
    {
      "@type": "Article",
      "headline": "How AI search chooses its sources",
      "author": { "@id": "https://www.example.com/#author" },
      "datePublished": "2026-10-01",
      "dateModified": "2026-10-20"
    }
  ]
}

Now every article by that author points to one Person, and that Person points to the profiles that confirm who they are. This is the same entity work that helps AI answers name you correctly.

Ship it where every crawler can read it

Most AI crawlers do not run JavaScript (see Can AI crawlers see your JavaScript?). Structured data added by a tag manager or a client-side script may be visible to Google after rendering and invisible to everyone else.

Write JSON-LD into the HTML template on the server.
Keep one source of truth for facts such as prices and dates, so the markup and the visible page cannot drift apart.
Use JSON-LD rather than microdata. It is easier to maintain and is the format Google recommends.

Check it

Rich Results Test (Google) shows which rich results a page is eligible for and flags errors.
Schema Markup Validator (schema.org) checks any type, including the many Google does not use for rich results.
Fetch the raw HTML with curl and confirm the JSON-LD block is in the response itself.
Search Console's enhancement reports show errors across the whole site once Google has crawled it.

Re-run the checks after template changes. A broken template can strip the markup from hundreds of pages at once, and nothing on the visible page will warn you.

Primary sources and further reading

Book-derived explanations on this page are editorial synthesis. Platform-specific claims link to the platform's own current documentation.

Frequently asked questions

Does Google use structured data for AI Overviews?

Google says structured data isn't required for its generative AI features and that there is no special schema to add. It still recommends structured data as part of normal SEO, and markup must match the visible content.

Do ChatGPT and Perplexity read schema?

Their crawlers download the HTML, which includes any JSON-LD in it. Neither company has published how much weight it gives markup, so treat it as useful context rather than a lever.

Should I remove FAQ markup now that FAQ rich results are gone?

No. FAQPage is still valid schema.org vocabulary. Keep it on pages that really are questions and answers, and remove it where it was added only to win the rich result.