Schema.org

Schema.org defines types and properties used to describe web content such as organizations, articles, datasets, and events. Publishers commonly express these descriptions as JSON-LD embedded in pages. Shared terms make a page easier for consuming systems to interpret, but the vocabulary is broad and deliberately does not impose all the validation rules a domain-specific record may need.

Structured markup is a publisher's assertion, not independent evidence that the page's claims are correct. Types can overlap, and a property can have several acceptable value forms. A consuming system should compare markup with visible content and authoritative sources before using it as a fact. Record the page URL, retrieval date, and exact markup when a downstream decision depends on it.

A fictional company page carries markup stating that it is an organization named Halden Therapeutics AG with a founding date, a number of employees and an address. A researcher compiling company facts reads the markup and finds a founding date that differs from the one in the company register.

The record keeps both. The markup is cited as the publisher's assertion with the page URL and retrieval date, and the register as the authority for the legal fact. The difference is reported as a conflict between a self-published description and a register, with the register preferred for the founding date under a stated rule, and the markup kept as evidence of how the company describes itself. If the markup disappears in a later redesign, the earlier observation remains in the record with its date.

Schema.org is not a data validation standard and not a registry. It does not say which values are correct, and passing a markup tester says that the syntax and the expected types are acceptable to that tool. Different consumers have different requirements, and a page that qualifies for one feature may not for another.

It is also not a replacement for a domain ontology. Schema.org terms are broad on purpose, and a regulated domain needs more precise vocabularies and controlled terms. A system of record can use Schema.org as an interchange layer for publication while keeping a richer internal model, and should say which statements are exported and which are not.

The vocabulary changes in releases that add types and properties, and occasionally mark terms as superseded or pending. Publishers who copy markup from templates may continue to use terms that have been superseded, and consumers vary in how they treat them. A record that stores extracted markup should keep the vocabulary release current at extraction time when it is known, and the retrieval date when it is not. Search engines also change which properties they use, so a markup property that mattered for one feature at the time of publication may have no effect later, and its presence is then a historical fact about the page rather than a signal of anything current.