What a machine does with your nouns
Ask an AI assistant who fixes flat roofs in Miami. To answer, it pulls fragments from pages, reviews, directories and structured data, and has to settle three questions about each one:
- Extraction. Can it pull your name, your services and your area out of the page without guessing?
- Disambiguation. Does "Acme Roofing" mean you, or the Acme Roofing in Ohio? (Entity resolution covers this.)
- Attribution. Which source does each fact come from?
An ontology helps with all three: named things, typed things, related things. Prose can say the same, but a reader has to infer it. Structured data states it.
What Google says it does with structured data
Google is explicit about two uses. Its introduction to structured data (last updated 10 December 2025) says Google uses structured data it finds on the web "to understand the content of the page, as well as to gather information about the web and the world in general, such as information about the people, books, or companies that are included in the markup". Its Organization guide (last updated 8 September 2026) says the url property "helps Google uniquely identify your organization".
Google also says what structured data does not do: it "does not guarantee that features that consume structured data will show up in search results".
How many websites publish structured data
Web Data Commons, a research project, extracts Microdata, JSON-LD, RDFa and Microformats from each Common Crawl web corpus. Its last three October releases:
| Common Crawl | HTML pages parsed | Pages with structured data | Websites in crawl | Websites with structured data |
|---|---|---|---|---|
| October 2022 | 3,048,746,652 | 1,518,609,988 | 33,820,102 | 14,235,035 |
| October 2023 | 3,353,090,410 | 1,696,953,312 | 34,144,094 | 14,646,081 |
| October 2024 | 2,391,039,772 | 1,245,622,627 | 37,447,141 | 16,525,070 |
"Websites" here are pay-level domains, which Web Data Commons counts without subdomains. Its summary for October 2024 gives 51.25% of pages and 44.12% of websites. The crawls differ in size and only partly overlap, and the project says trends across them "must be interpreted with caution".
Which format and which types, October 2024
| Format | Websites using it | Pages using it |
|---|---|---|
| JSON-LD | 11,562,359 | 833,818,654 |
| Microdata | 7,599,792 | 574,648,578 |
| Microformat hCard | 3,522,517 | 179,698,044 |
| RDFa | 474,635 | 49,636,704 |
| Schema.org type in JSON-LD | Websites using it |
|---|---|
| WebSite | 8,461,613 |
| Organization | 6,574,284 |
| WebPage | 6,309,173 |
| BreadcrumbList | 6,206,556 |
| Person | 4,726,830 |
| Product | 2,349,844 |
| PostalAddress | 1,971,988 |
| LocalBusiness | 1,263,114 |
Schema.org's own home page gives a broader figure from a different count: "As of 2024, over 45 million web domains markup their web pages with over 450 billion Schema.org objects."
What the numbers mean for you
Organization markup is on millions of websites, so having it is common. The ontology work is in the connections: one @id for the company, the same name everywhere, and relationships that say who offers what, where. A list of disconnected facts leaves the reader to guess how they fit together.
Sources
All checked 13 November 2026:
- Web Data Commons, Microdata, RDFa, JSON-LD, and Microformat Data Sets and October 2024 statistics
- Google Search Central, Introduction to structured data markup and Organization structured data
- Schema.org, home page, version 30.1
Next step
Start building: write your first triples.
Frequently asked questions
- How many websites publish structured data?
- In the October 2024 Common Crawl, Web Data Commons found structured data on pages from 16,525,070 of 37,447,141 pay-level domains, which it reports as 44.12%. JSON-LD was used on 11,562,359 of those domains and Microdata on 7,599,792.
- Does publishing structured data guarantee a rich result or an AI mention?
- No. Google says it does not guarantee that features that consume structured data will show up in search results. Structured data makes your facts easier to read; it does not decide what a system shows.