In plain terms
A person who sees “1,250 TL” next to a photo of a chair understands that it is a price. A machine sees a string of characters. Structured data states the same facts a second time, in a vocabulary that search engines have agreed on, so that software can read them without guessing. Visitors never see it.
Why it matters
It is how search results come to show prices, ratings, opening hours and event dates, and one of the ways search engines learn which company, person or product a page is about. For AI search the honest position is this: search engines say it helps them understand pages, while evidence that it directly raises citations in AI answers is limited. It is cheap, well established and harmless when it matches what is on the page, so it belongs in the basics.
Example
Every page of this glossary says, in a small block of code, that it describes a defined term, and gives the term's name, its description and the glossary it belongs to. A recipe site does the same with cooking time and ingredients, and an online shop with price and stock. That is why a search result can show a price or a cooking time before anyone clicks.
Most often confused with
Structured Data vs. Structured output
The names are close and both often involve JSON, but they sit at opposite ends. Structured data is something a website publishes about its own content. Structured output is something a language model produces so that other software can process its answer.
Origin: Schema.org was launched in 2011 by Google, Microsoft and Yahoo, with Yandex joining soon after.
Under the hood
Schema.org is the shared vocabulary: several hundred types (Organization, Product, Offer, Article, FAQPage, LocalBusiness, Event, Person, DefinedTerm and others), each with defined properties. Three formats can carry it: JSON-LD, which is a script block separate from the visible HTML and the format Google recommends; Microdata; and RDFa. Rules: the markup must describe content that is really visible on the page, and misleading markup can lead to penalties. Marking up a page makes it eligible for enhanced results and does not guarantee them; search engines have added and retired result types over the years. Properties such as sameAs connect an entity to its other profiles and help knowledge graphs identify it. Validate with the Schema Markup Validator and Google's Rich Results Test. JSON-LD sits in the raw HTML, so crawlers that do not run JavaScript can still read it.