- [Home](https://www.agenticsearch.at/en/) - [Knowledge](https://www.agenticsearch.at/en/knowledge/) - Schema markup for AI # Schema markup for AI systems ## Structure removes room for interpretation. Structured data is not a ranking lever and still a precondition. It tells a system unambiguously what plays which role on a page.  Principle ## One graph instead of many snippets. Most websites carry structured data as separate blocks: an Organization snippet here, an Article there, an address on the contact page. Each correct on its own and connected to nothing. **Better is one connected graph with stable anchors.** Every page contributes a slice of the same model, and the nodes reference each other through fixed identifiers. Then a system knows the organisation on page A is the one on page B. In practice that means fixed anchors such as `#organization`, `#website` and `#person` that read the same on every page. We do it that way on this website, visible in the source of any page you like. Relevant ## Which types actually count. ### Organization The core entity. Name, legal form, address, register, sameAs. The anchor everything else points at. ### Person For authorship and responsibility. What matters is mainEntityOfPage, so it is clear where the main entity lives. ### WebPage and Article Connects the address to the entity and carries the modification date. For articles, authorship on top. ### FAQPage Removes any doubt about which paragraph answers which question. Has to match the visible text. ### BreadcrumbList Shows the position in the structure. Should come from the same source as the visible breadcrumb. ### Product and Offer Decisive for shops. Price, currency, tax and availability have to match the page. Limits ## What markup cannot do. Markup does not turn a thin page into a good one. It describes what is there, and where little is there it describes little. Markup also replaces no mentions outside your own domain. A perfectly marked up self-description stays a self-description. And markup that contradicts the visible text is worse than none. If the markup carries one price and the page another, you produce exactly the kind of contradiction that leads to wrong answers. E-E-A-T, incidentally, is not markup. There is no property for experience or trustworthiness. What exists are statements about authorship and sources, and those can be marked up. Common questions ## About structured data. **JSON-LD or microdata?** JSON-LD. It sits separately from the markup, can be generated from one source and is easier to test. Microdata spreads the same information across the HTML. **How many types should be marked up?** As many as apply and no more. Types that describe nothing inflate the graph without clarifying anything. **What are stable anchors?** Fixed identifiers such as yourdomain.at/#organization that read the same on every page. They let you point at the same node from anywhere. **Does the markup have to match the visible text?** Strictly. Deviations breach the guidelines and produce contradictions that show up in answers. **Does AggregateRating help?** For products and businesses yes, when the reviews are real and attributable. Self-issued ratings with nothing behind them are a risk. **Does author markup count?** For attributing statements to people yes. It does not replace a verifiable person with traces outside your own website. **How do you test the graph?** With the search engines' testing tools plus a check of your own on whether every internal reference resolves. We do the latter during the build. **Does every page need the same graph?** The same core, plus the page-specific part. Organisation and website repeat, the page type and the breadcrumb do not. **What about speakable?** The type exists and is barely evaluated. We do not use it, because we could not observe any effect. **Where do you start when nothing exists?** With Organization carrying correct company details and with WebPage per page. That is the base everything else builds on. Next step ## Let us find out where you stand in AI answers. One conversation, 30 minutes, no sales pressure. You leave with an honest read on whether this is worth the effort for you. [Look at the audit first](https://www.agenticsearch.at/en/services/ai-visibility-audit/) [Request a call](https://www.agenticsearch.at/en/contact/)