AI Visibility pillar / check schema-completeness-for-ai

Page-type structured data

Whether a page carries the structured data that matches what the page actually is, and whether that markup matches what visitors can see. Google requires no special schema for its AI features.

By Shimon Carroll, Founder, SEO for AI Agents · Last updated

What this check measures

The check classifies the page from its URL and visible signals: the site root, an article (a blog, news, guide, or resources path, or a long page wrapped in an article element), a product page (a product or shop path with visible purchase signals), or any other page. It then expects only the structured data that fits: an Organization or a LocalBusiness subtype on the homepage, Article (or BlogPosting or NewsArticle) on articles, Product on product pages, and BreadcrumbList on pages two or more path segments deep. Separately, it compares every FAQ question in the markup with the visible page text.

Why it matters

Google states that structured data is not required for generative AI search and that there is no special schema.org markup to add for AI Overviews or AI Mode. Structured data is still a useful part of ordinary SEO: it helps search engines understand who runs a site and what a page is, and it is how pages qualify for rich results. The rules that matter are that the markup is a true representation of the page and never describes content readers cannot see. Asking every page for every type produced markup that did not match the page, which is the opposite of what the guidelines ask for.

How we score it

FAIL (medium) when the homepage has no Organization or LocalBusiness-type entity. FAIL (low) when an article, product, or deep page lacks its matching type. FAIL (medium) when FAQ markup contains questions that are not visible on the page. PASS when the page carries what its type calls for, or when nothing on the page calls for a specific type. A finding counts once per audit however many pages share it.

Confidence-flag rules

Confidence is MEDIUM: the page-type classification is a heuristic built from the URL and visible signals, so an unusual URL structure can be misread. The evidence lists the detected page type, the types found, and the types expected, so the classification can be checked.

Common mistakes

  • Adding FAQPage or HowTo markup to pages that have no visible FAQs or steps.
  • Marking up questions, reviews, or prices that are not shown to visitors.
  • Treating structured data as an AI ranking switch. Google says no special markup is needed for its AI features.
  • Shipping JSON-LD only through client-side JavaScript, where some crawlers never see it.

How to fix it

Add the one type the finding names, describing only what is visible on that page: an Organization (or your LocalBusiness subtype) with your real name, URL, and logo on the homepage; Article with the visible headline, author, and dates on articles; Product with the visible name and price on product pages; BreadcrumbList that mirrors the visible breadcrumb trail on deep pages. Validate with the Schema Markup Validator or the Rich Results Test, and render the JSON-LD on the server.

Primary sources

Changelog

  • · Initial publication. The check looked for a fixed set of eight types (Organization, WebSite, Article, FAQPage, HowTo, BreadcrumbList, Product, Service) on every page.
  • · Rewritten to be page-type-specific. A page is no longer asked for types that do not describe it (Product on a contact page, HowTo on a homepage). FAQPage and HowTo are never expected. Added a check that FAQ markup matches visible content. Default severity lowered from high to medium, and the copy now states Google guidance plainly: no special schema is required for AI Overviews or AI Mode. Earlier claims that sparse schema causes AI engines to demote a site were removed because no first-party source supports them.
← Back to the methodology indexRun an audit