Schema markup checker

Every <script type="application/ld+json"> block in the delivered HTML, in document order. For each block: whether it is valid JSON, what @context it declares, whether it uses @graph, and every entity inside it — including entities nested inside another entity’s properties, which is where most of a page’s structured data actually lives.

For each entity it reports the @type, the @id, the properties present, and — for the types the bundled ruleset covers — which expected properties are absent. Duplicate @id values across entities are called out, because two entities sharing an identifier are read as one and have their properties merged.

Ruleset version 2026-08, bundled with SearchOrbit AI: Organization, LocalBusiness, WebSite, WebPage, Article, Product, BreadcrumbList, FAQPage, SoftwareApplication, VideoObject, Event, JobPosting.

An entity of any other type is still reported — its @type, its @id and its properties — and simply marked as not checked. A type this ruleset does not cover is not a type that is wrong.

These findings come from SearchOrbit AI’s own ruleset, and the version that produced them is stamped on every result. Google’s requirements for a given rich result change without notice, and no third-party tool can promise otherwise — a checker that presented its own snapshot as an official answer would be wrong the first time a requirement moved and silent about it afterwards.

Nothing here says a page is eligible for a rich result. That is Google’s decision, and the Rich Results Test is where to confirm it.

• ·Nothing is executed. JSON-LD is parsed as data — there is no eval, no expansion step, and no way for markup on a page to run anything here.
• ·A remote @context is never fetched. It is recorded as text and compared against schema.org; terms defined only in that context are reported as unchecked rather than resolved.
• ·No URL inside the markup is requested. An image, a sameAs, a logo, an @id — all are reported as declared values, so a broken one is not reported as broken.
• ·JSON-LD only. Microdata and RDFa are detected and reported when present, and explicitly not validated.
• ·It reads the HTML as delivered. Structured data a page injects with JavaScript after loading is not in that HTML and is not reported.
• ·Pasted markup is limited to 128 KB, is never stored and is never cached.
• ·No score is invented. Findings are counted by severity, because a number nobody documented invites comparisons it cannot support.

Structured data is one signal among several. To see the rest of what a crawler reads on the same page, run a free SEO audit or check its meta tags.