Structured data for LLMs is a broad term for publishing information in a machine-readable format that an AI system or its retrieval tools can use. On the web, one approach is to describe a page’s entities and relationships with the Schema.org vocabulary in JSON-LD.
What can structured data describe?
Schema.org defines types and properties. JSON-LD is a format for expressing that information. A glossary page, for example, can identify its entry as a DefinedTerm, give it a name and description, and link it to a DefinedTermSet. The description should agree with the visible definition. Whether a particular system reads or uses those fields depends on how that system retrieves and processes pages.
How is it different from semantic HTML?
JSON-LD can describe what a page is about. Semantic HTML describes the structure and controls that appear on the page. For example, the glossary entry’s JSON-LD can identify a term, while a native HTML button gives the browser the semantics for a control that opens a panel. Both can be useful on the same page.
Do Google’s AI features need special structured data?
Google says its AI Overviews and AI Mode require no special Schema.org markup. A page needs to be indexed and eligible to appear with a snippet to be eligible as a supporting link. Whether a page appears as a source depends on the system and the query. Keep your visible content clear and use structured data to describe it faithfully.