Guide
What is structured data?
Machine-readable facts added to a page so nothing has to guess what it is about. Well understood for search engines, and genuinely unsettled for AI ones, which most pages on this subject will not tell you.
What it actually is
A page says what it is twice. Once in prose, for a person, and once in a block of machine-readable facts, for everything else. That second copy is the structured data. It names the thing the page is about, gives it a type from a shared vocabulary, and lists its properties: a name, an address, a price, an author, a date.
The point is removing interpretation. A person reading “open from nine” knows what that means. A machine has to decide whether that is an opening time, a quotation or part of a sentence about something else. Structured data says it is an opening time, in a field called opening hours, and the question does not arise.
Three words people use interchangeably
They are related and they are not synonyms, and the confusion makes the subject harder than it is.
Structured data is the idea: facts in a fixed shape. Schema.org is the vocabulary nearly everyone uses, a shared list of types and properties maintained jointly by the major search engines, so that Organization means the same thing to all of them. JSON-LD is the file format those facts are written in, a block of JSON in a script tag. There are older formats, microdata and RDFa, which wrap the visible HTML instead. JSON-LD sits apart from it, which is why it won.
What it reliably does
This part is documented and not in dispute. Search engines use structured data to understand what a page is about and to decide whether it qualifies for a richer result: review stars, an FAQ expander, a recipe card, an event listing. Google publishes which types are eligible and tests them in its own tools.
It is not a ranking factor, and that distinction is worth holding on to, because almost every tool that generates a schema report implies otherwise. Correct markup changes what your result can look like and what it is eligible for. It does not move you up the page.
What the evidence says about AI engines
Here the honest answer is that nobody knows, and the pages telling you otherwise are mostly selling something. No engine vendor documents using schema.org markup to decide citations. What exists instead is a set of vendor studies pointing in both directions, several of them reporting the same secondhand experiment, and figures such as an eighty-nine per cent correlation that come with no published method attached.
What can be reasoned about is the mechanism, because the engines are not built the same way. An engine answering from a search index inherits whatever that index already extracted, markup included. An engine that fetches your page live at the moment of the question is reading the page as delivered, and some of those crawlers convert what they fetch to plain Markdown first, which throws away a JSON-LD block before the model ever sees it. Those two paths cannot plausibly give the same answer, so a single confident claim about “AI engines” is already suspect.
The practical position this leaves is unglamorous and holds either way. Markup is cheap to add, costs nothing when ignored, and removes ambiguity for anything that does read it. So add it. Just do not let it replace the work of saying plainly, in the visible words on the page, what your business is and what it sells. Every engine reads that.
What to put on a business site
Identity first, because it is the one a model most often has to guess at. An Organization or LocalBusiness block on the home page, carrying the legal name, the address, the contact details and the links to your profiles elsewhere. That single block answers the question underneath most of what an engine wants to know about you.
After that, mark up what the page genuinely shows. Product or Service on the pages that sell, Article with an author and a date on anything written, FAQPage where real questions and answers appear on the page. The rule that keeps you out of trouble is that the markup describes what a visitor can see. Markup that contradicts the visible page is the one version of this with a penalty attached.
How to check what you have
Most sites already carry some, added by a plugin or a theme years ago and never looked at since. The failure worth finding is the silent one: a syntax error that stops the whole block being read, which changes nothing visible and so survives indefinitely.
Our structured data checker reads every JSON-LD block on a page, says which fail and why, lists the types you carry and gives you filled-in stubs for the ones a business page usually needs. It is free and needs no account.
Questions about structured data
What is structured data?
A block of machine-readable facts added to a web page, saying in a fixed vocabulary what the page is about: that this is a company, that its name is this, that it sells these things, that it is based here. The visible page says the same in prose. The structured data says it in a form nothing has to interpret.
Is structured data the same as schema markup?
Near enough in everyday use, and they are not the same word. Structured data is the general idea of machine-readable facts on a page. Schema.org is the shared vocabulary almost everyone uses for it. Schema markup means structured data written in that vocabulary. If somebody says one and means the other, nothing is lost.
What is JSON-LD?
The format the facts are written in. A script tag holding a block of JSON, sitting in the page source separately from the visible HTML rather than wrapped around it. It is the format Google recommends, and it won because you can add it without touching your markup or your layout.
Which pages on a website need structured data?
The ones that make a claim worth confirming. Your home page and about page carry who you are. Product and service pages carry what you sell. Articles carry who wrote them and when. A page that exists only to link elsewhere needs nothing. Marking up every page because you can produces maintenance work and no benefit.
Is structured data a ranking factor?
No, and the people selling it rarely say so plainly. Google has been consistent that it is used to understand a page and to decide eligibility for rich results, not to rank it higher. Correct markup can change how your result looks and whether it qualifies for certain formats. It does not move you up the page.
What happens if my structured data is wrong?
Usually nothing visible, which is why broken markup survives for years. The common failures are a syntax error that stops the block being read at all, and markup describing something the page does not show. The second is the one with a penalty attached: Google treats markup that contradicts the visible page as spam, and can drop the site from rich results entirely.
Markup is one of five things an engine needs.
Crawler access, structure, whether a model can say what you do, speed and security. Scored together, free, in under a minute.