Structured data is information about a page written in a fixed, machine readable vocabulary instead of ordinary sentences. It exists because software reads strings of text poorly and reads labeled facts well. The bottom line: AI systems and search engines use structured data mainly to work out which real world entity a page is about and what its stated facts are, and the evidence supports that it improves machine accuracy, while the evidence that it raises citations is thin.
See also: answer engine optimization, explained step by step.
From strings to things
For its first decade, search matched the words you typed against the words on pages. In May 2012 Google announced the Knowledge Graph with the phrase "things, not strings," describing a database of 500 million objects and 3.5 billion facts about them and the relationships between them. By May 2020 the figure had grown to about 500 billion facts on 5 billion entities, and Google says the graph draws on many sources, including Wikipedia. The point of the change was disambiguation: a page about Jaguar could be about a car, an animal, or a football team, and a graph of entities lets a system decide which.
The vocabulary and the formats
In 2011, Google, Microsoft, Yahoo, and Yandex launched schema.org, a shared vocabulary of types such as Organization, Product, and Article and their properties. A page can express that vocabulary in three formats: Microdata and RDFa, which are woven into the visible HTML, and JSON-LD, which sits in one block. JSON-LD became a W3C Recommendation in January 2014 and is the format Google recommends because it can be added and maintained without touching the visible markup.
What a machine sees, side by side
| The page says in plain text | The same fact as structured data |
|---|---|
| Open every day until 9 | openingHoursSpecification with dayOfWeek, opens, and closes values |
| Serving the Tampa Bay area | areaServed listing Tampa, St. Petersburg, and Clearwater |
| Call us at 555 0142 | telephone with a full international number |
| Follow us on LinkedIn | sameAs linking to the exact company profile URL |
A person reads all four correctly. A machine reading plain text has to guess whether "until 9" means morning or evening, which days, and which time zone. The structured version removes the guesswork.
Why AI systems care
Language models answer questions by combining what they learned in training with material retrieved at question time. The more precisely a retrieved fact is labeled, the less the system has to infer. A data.world benchmark by Sequeda, Allemang, and Jacob shows the size of that effect in a controlled setting: GPT-4 answered enterprise questions with 16.7 percent accuracy against a raw SQL database and 54.2 percent when the same data was expressed as a knowledge graph. Microsoft's Fabrice Canel said in March 2025 that schema markup helps Microsoft's language models understand content for Copilot. Google's guidance for AI Overviews and AI Mode adds that structured data should match the visible text and that Merchant Center and Business Profile information should be kept up to date, which are structured feeds Google already reads.
What it does not do
Google states that no special schema is required to appear in AI Overviews or AI Mode. A December 2024 Search/Atlas study found no correlation between schema coverage and citation rates, and an Ahrefs controlled test of 1,885 pages found no meaningful lift, although those pages were already heavily cited. Structured data helps a machine understand you. Whether it then chooses you depends on retrieval, relevance, and corroboration.
How to use this
- Add the most specific business type for your organization, with address, phone, hours, and service area.
- Link your other profiles with sameAs.
- Mark up blog posts with Article and deep pages with BreadcrumbList.
- Keep every value identical to what the page shows.
- Check that the markup is present in the raw HTML source.
Questions people ask
What is structured data in simple terms?
It is information about a page written in a fixed vocabulary, such as schema.org, so software can read labeled facts like a business's hours or address instead of guessing from sentences.
What is the difference between structured data and JSON-LD?
Structured data is the concept. JSON-LD is one format for writing it, a block of JSON in the page. Google recommends JSON-LD over Microdata and RDFa.
Do AI systems use structured data?
Microsoft says schema helps its language models understand content for Copilot. Google says no special schema is needed for AI Overviews. No published study shows schema raises citation rates.
Where does structured data come from historically?
Google announced its Knowledge Graph in May 2012, schema.org launched in 2011, and JSON-LD became a W3C Recommendation in January 2014.