Schema.org
Schema.org is a shared vocabulary for structured data: a set of types and properties you use to label the facts on a page — a price, a rating, an opening hour — so a machine doesn't have to guess what a string of text means. Google, Bing, Yahoo and Yandex founded it in 2011, and it's still the closest thing the web has to a common markup standard. It makes your content easier to parse. It does not make it more likely to be cited.
Why Schema.org matters for AI visibility
Be precise about what schema actually buys you, because the claims made about it online are often overstated. Google's own guidance on AI Overviews and AI Mode says no special markup, schema, or AI-specific file is required to appear in those surfaces, and schema has not been a confirmed ranking factor since 2018. What structured data does do is remove ambiguity: it hands a crawler or model a name, address or price already labeled, instead of leaving it to infer one from prose. That's a machine-readability improvement, not a proven citation lever. The stronger, evidence-backed lever is being talked about elsewhere — Ahrefs' analysis of roughly 75,000 brands found web-mention frequency correlates with AI citation rate at about 0.664, roughly three times the correlation for backlinks (about 0.218). Schema is still worth doing correctly; just don't expect it to substitute for being mentioned.
How Schema.org works technically
Schema.org itself is only the vocabulary — the list of permitted types and properties. A type is, for example, LocalBusiness, Product or FAQPage; a property on that type is something like name, address or priceCurrency. To make these details machine-readable you need a serialization format. The one recommended today is JSON-LD, a self-contained data block in the page's source that stores the facts separately from the visible text. Older formats, microdata and RDFa, are woven directly into the HTML instead. The rule that matters most: whatever you put in the markup must match what a visitor actually sees on the page. Ratings, prices or availability that exist only in the data block and not in the visible content are treated as spam and can get the page penalized.
Common mistakes
The costliest mistake is mismatch between markup and visible content — five-star ratings in the JSON-LD with nothing to back them up on the page. Stale data is nearly as common: opening hours or a business name updated in the running text but never touched in the schema block. Plenty of sites also pick the wrong type, tagging a service as a Product, or leave out required properties, which leads engines to ignore the markup outright. Test your markup regularly with Google's Rich Results Test or the Schema Markup Validator. A small amount of correct markup beats a large amount of broken markup. And treat schema as description, not decoration: it should describe what's already true on the page, not make a claim that isn't.
Example
Picture a small dental practice in Leipzig. The visible page lists its name, address, phone number and hours in plain text — obvious to a human reader, but just characters to a crawler. Add a JSON-LD block with the type LocalBusiness, with name, address, telephone and openingHours each labeled as such, and a machine no longer has to guess which string is which. That doesn't guarantee an AI assistant will recommend the practice. It does mean that if it does pull from the page, it's less likely to quote the wrong phone number or yesterday's hours.
Common questions
Does adding Schema.org markup boost my AI citation rate?
Not on its own. Google states plainly that no special schema is required for AI Overviews or AI Mode, and schema stopped being a direct ranking factor in 2018. It improves machine-readability, which can help a model extract your facts correctly once it's already looking at your page — it isn't what gets you cited in the first place.
Which format should I use to implement it?
JSON-LD. It's the format Google recommends, it lives as a separate block in the source code, and you can maintain it without touching your visible HTML. Microdata and RDFa still work but require markup inline in the page and are more fragile to keep in sync.