Glossary / Structured data for AI (schema markup)
Structured data for AI (schema markup)
Also: Schema markup, JSON-LD, Schema.org markup
Structured data is schema.org markup, usually written as JSON-LD, that labels a page's facts in a machine-readable format so AI systems can extract and attribute them without having to interpret the surrounding prose.
Structured data is a block of machine-readable markup, almost always written as JSON-LD in a script tag, that states a page's facts explicitly instead of leaving them buried in sentences. It follows the shared vocabulary at schema.org: each block declares a @type such as Organization, Product, Article, or FAQPage, and carries the properties that type expects, for example a product's name and price, or an article's headline and author.
Prose describes; schema declares. A paragraph might mention a price inside a sentence a machine has to parse and infer, while the same price inside Product schema is a labeled field with no interpretation required. That clarity only helps if the markup actually reaches the reader, and most AI crawlers fetch raw HTML without running JavaScript, so structured data injected client-side never arrives.
What structured data is
Structured data uses the schema.org vocabulary, a shared set of types and properties that Google, Bing, and other consumers agreed on. The dominant format is JSON-LD, a block of JSON placed in a script tag of type application/ld+json and kept separate from the visible markup so it does not affect layout. Older formats, microdata and RDFa, interleave attributes into HTML tags instead, and JSON-LD is preferred because it is self-contained and easy to validate.
Each block declares one or more entities. An entity has an @type (Organization, Product, Article, FAQPage, BreadcrumbList, and hundreds more) and a set of properties. The type dictates which properties carry meaning: an Organization is identified by its name and url, a Product by at least its name, an Article by its headline, an FAQPage by the questions and answers in its mainEntity. Declaring a type together with the properties it expects is what turns a block of text into a fact a machine can lift out cleanly.
Why it matters for AI visibility
AI systems build answers by extracting facts and attributing them to a source, and structured data makes both steps less error-prone. When your organization name, author, publish date, product price, or the exact question a page answers is labeled rather than inferred, an AI system is less likely to misattribute it, confuse it with a competitor, or state it wrong.
Two honest limits matter. First, structured data is not a lever you can pull for guaranteed citations. To be cited by an AI system a page must first be readable, allowed, and structured, and schema strengthens a source that already clears those bars rather than creating selection where the fundamentals are missing. Second, schema only counts if the crawler receives it. Because most AI crawlers do not execute JavaScript, markup added by a client-side script or a tag manager reaches them as nothing. Structured data has to sit in the raw HTML the server sends, next to the content it describes.
How Lantad measures and improves it
Schema is one of the five dimensions in Lantad's AI Visibility Score. Lantad reads the raw HTML that an AI crawler receives, the same fetch it uses for Prose Parity, and checks the JSON-LD three ways: that it parses as valid JSON, that each block declares a @type, and that recognized business types carry their required properties, for example Organization with name and url, or FAQPage with mainEntity.
Two rules keep the score honest. Structured data that appears only after JavaScript runs is discounted, because crawlers that do not run JS never see it, and the report shows which blocks fell into that gap. A page whose only markup is a generic type like WebPage carries structured data but no business fact to validate, so it does not earn full marks. Schema is one input to the composite, so a page that is perfectly readable, allowed, and structured but ships no structured data still leaves points on the table: readable is not the same as complete.
Common questions
- Does schema markup help with AI search and ChatGPT?
- It helps machines extract and attribute your facts accurately, which is a prerequisite for being used as a source. Schema does not guarantee a citation, and it only counts if it sits in the raw HTML an AI crawler receives, since most AI crawlers do not run JavaScript. Labeled facts are easier to lift and less likely to be misstated than facts a system has to infer from prose.
- Which schema types matter most for AI visibility?
- The types that state a verifiable business fact: Organization (who publishes the page), Article and its subtypes (headline, author, date), Product (name and price), FAQPage (the exact questions a page answers), and BreadcrumbList (where the page sits). Lantad recognizes these and checks that each carries its required properties. A generic WebPage type adds no fact a machine can validate.
- Does structured data guarantee my page gets cited by AI?
- No. Being read is a prerequisite for being cited, never a guarantee of it. Structured data makes your facts easier to extract and attribute, but AI systems choose sources on factors no vendor fully publishes. The reliable path is to be readable, allowed, and structured first, then let schema make the facts you already publish unambiguous.
See if AI can read your site
Paste a URL and see exactly what AI crawlers can and cannot read, with a graded report and ranked fixes. Free, no signup.