What this check tests
For each crawled page the audit parses every JSON-LD block in the served HTML (see Schema.org JSON-LD present for how pages are chosen and blocks parsed). A page counts when one of its blocks has a top-level @type that is a single string containing, as a plain substring, any of these 16 names:
WebPage, Article, BlogPosting, Product, LocalBusiness, Organization, Person, FAQPage, HowTo, Recipe, Event, Course, SoftwareApplication, WebSite, ItemList, BreadcrumbList.
The first matching block on a page is enough. The score is the share of all crawled pages that count, including pages with no JSON-LD at all.
- Pass: 50% or more of crawled pages count.
- Partial: at least one page counts but fewer than half.
- Fail: no page counts.
Four behaviours decide most results.
- The match is on the substring, so
NewsArticleandTechArticlecount throughArticle, andProductGroupcounts throughProduct. The match is case-sensitive. - A type outside the list does not count, even when it is the right one:
CollectionPage,AboutPage,ContactPage,Service,DatasetandVideoObjectare examples. - Only a string
@typeat the top level of a block is read. A block whose@typeis an array, or a block that wraps its nodes in@graphwith no top-level@type, is skipped. Some generators emit@graph, so a site can have rich markup and still score low here. - The check cannot tell whether the type fits the page. A
Productblock on a blog post counts.
Why it matters to agents
The type tells a parser what kind of thing the page describes. Schema.org arranges types in a hierarchy (NewsArticle sits under Article, which sits under CreativeWork, under Thing), and says a type may have several parents. A consumer that knows only the general type can still use what the page says about it. That is the case for choosing the most specific type that is true.
For Google, the type decides which feature guide applies. Its guides cover types such as Article, Event, Course, Dataset, Local business and Organization, each with its own required and recommended properties. A type outside those guides is still valid schema.org, but Google documents no feature for it.
This is a weak, listed-names check, not a judgement of quality. Its weight is 1.
How to fix it
- Pick the most specific schema.org type that is true for the page. A blog post is
BlogPostingorArticle, a company page isOrganizationorLocalBusiness, a pricing or product page isProductorSoftwareApplication, a question list isFAQPage. UseWebPageonly when nothing narrower applies. - Put that type in a single string
@typeon the top-level object, as in"@type": "Article". If your plugin builds a@graph, confirm that the page's main node is also output as its own block with a top-level string@type, or add one. - View the page source and confirm the block is in the served HTML, then check the type spelling and capitalisation.
articledoes not matchArticle.
If the type you need is not in the list, the check will not count it. The check is a heuristic, so do not swap a correct Dataset for a wrong WebPage to move a score. Add a WebPage or Article block alongside it if that is also true of the page.
Next, Schema.org completeness checks which fields the first block carries, and JSON-LD valid checks @context and @type. For background see Schema.org for AI.