Structured data is information added to a web page's code in a standardized format, so search engines and other programs can read facts about the page (what it is, who made it, what it costs) without interpreting the visible text. Most structured data on the web uses the Schema.org vocabulary and is written in a format called JSON-LD.
How does structured data work?
A page written for people mixes facts with layout, navigation and prose. A crawler, the program a search engine uses to fetch pages, can read every word but has to guess which number is a price and which is a rating. Structured data removes the guess by labelling each fact with an agreed name.
The labels come from a vocabulary, a shared list of types and properties. Schema.org is the vocabulary the major search engines created together and still maintain. It defines types such as Product, Article, Organization, LocalBusiness, Event and Recipe, and properties for each type, such as name, price, author or openingHours.
A typical block for a product page looks like this:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Product",
"name": "Trail Running Shoe",
"brand": { "@type": "Brand", "name": "Example Outdoor" },
"offers": {
"@type": "Offer",
"price": "120.00",
"priceCurrency": "EUR",
"availability": "https://schema.org/InStock"
}
}
</script>The block sits inside the page's HTML, where visitors never see it. A crawler that fetches the page reads it together with the visible content and stores the labelled facts in its index.
What formats can structured data use?
Schema.org types can be written in three formats. All of them carry the same information; the difference is where the labels live.
| Format | Where it lives | Practical notes |
|---|---|---|
| JSON-LD | A separate script block, usually in the page head | Easiest to add and edit; the format Google recommends |
| Microdata | Attributes added to existing HTML tags | Labels sit on the visible elements, so edits touch the page template |
| RDFa | Attributes added to existing HTML tags | Similar to Microdata; common in some publishing systems |
Search engines accept all three, and a page can mix them. Most sites use JSON-LD because a developer or a content management plugin can generate the block without changing the visible layout.
What does structured data do in search results?
The most visible effect is rich results: search listings with extra elements such as review stars, prices and stock status, recipe cooking times, event dates or breadcrumb trails. Markup makes a page eligible for these features, but the search engine still decides whether to show them for a given query.
Structured data also helps a search engine connect a page to an entity, meaning a specific thing such as a company, a person or a product. Organization markup on a homepage, with the official name, logo and sameAs links to the brand's profiles on other sites, tells the engine that these pages describe the same company.
Search engines publish rules for markup. It has to describe content that is visible on the page, and the values have to be accurate. Markup that describes hidden content or invents reviews can lead to a manual action, a penalty in which a reviewer removes the site's rich results.
How do you check structured data?
You can check markup on a single page and across the whole site.
- Validate the syntax. The Schema Markup Validator, run by Schema.org, reads any markup and reports missing brackets, unknown types and misspelled properties.
- Test eligibility. Google's Rich Results Test shows which rich results a single URL qualifies for and which required properties are missing.
- Monitor the site. In Google Search Console, the enhancement reports list pages with valid markup, warnings and errors for each supported type.
How the markup reaches the page also matters. If a script adds the JSON-LD after the page loads, only crawlers that execute JavaScript will see it. Google renders JavaScript, but several AI crawlers read only the raw HTML the server sends. Putting the block in the server-rendered HTML makes it readable by every bot.
Does structured data matter for AI search?
The connection is less direct than with rich results. AI search systems retrieve pages from an index, clean them, split the text into passages and give the best passages to a language model, a process described in how AI search chooses which sources to cite. The cleaning step often removes script blocks, so a JSON-LD block is rarely the text the model reads or quotes.
Any influence comes through the index. Products such as Google AI Overviews are built on a search index that uses structured data to classify pages and link them to entities. When the index knows a page is a product page for a specific brand, the page is easier to match to questions about that product. AI companies have not published detailed documentation of how each system uses markup, so the size of this effect is unknown.
Structured data is sometimes compared with llms.txt, since both are summaries meant for machines. llms.txt is one Markdown file describing a whole site and is still a proposal. Structured data is attached to individual pages and is an established standard that search engines document and support.
What does structured data mean for a website?
Start with markup that matches what each page actually is: Organization on the homepage, Product on product pages, Article on blog posts, LocalBusiness for physical locations and BreadcrumbList for the site hierarchy. Every value in the markup should match the visible page. A price in JSON-LD that differs from the displayed price is a common error and gives search engines two conflicting facts.
For AI answers, the visible text still carries most of the weight, because passages are what the model quotes. Markup helps by making the page easier to classify. To learn whether adding or fixing it changes how often a brand appears in AI answers, track AI visibility for the same prompts before and after the change, over several weeks.
