Structured data for AI search: what it does, what it doesn't, and how I approach it

What the search engines have said

Google. Its guidance on AI features says the same foundational SEO practices apply to AI Overviews and AI Mode, including making sure structured data matches the visible content on the page¹. Google's general documentation describes structured data as a way to help it understand page content and to qualify for rich results². It doesn't promise AI citations in return.

Microsoft. Search Engine Land reported that Bing's Fabrice Canel said at SMX Munich in March 2025 that schema markup helps Microsoft's LLMs understand content, and suggested IndexNow for getting fresh content discovered³⁵.

Others. The same report noted it was unclear whether other AI engines use schema at all³. Claims of large citation lifts from markup alone mostly come from vendors and are hard to verify.

Think of it as a small ontology

Schema.org is a shared vocabulary of types and properties⁴, and JSON-LD on your pages forms a small knowledge graph about you. The same principles I use in enterprise ontology work apply:

• Stable identities. Give the person or organisation a single identifier and reference it from every page, so authorship and services all point to the same entity.
• Explicit relationships. Articles have an author; services have a provider; pages sit in a breadcrumb hierarchy. Say so in the markup.
• Consistency with what people see. Markup that claims something the page doesn't show is a data quality failure, and Google's guidance says to keep them matched¹.
• Validation. Check markup on every build, just as I would validate a knowledge graph with SHACL shapes.

What I put on a professional services site

Alongside markup, a few unglamorous things matter more than any schema type: fast pages, clear headings, accurate dates, a sitemap, and pushing new URLs to search engines promptly. This site uses IndexNow, an open protocol for notifying participating search engines of changes⁵, triggered whenever the sitemap changes.

What actually earns AI citations

Nobody outside the AI companies knows exactly how their systems choose sources, and anyone who claims certainty is selling something. What is reasonable to expect: content that answers a specific question clearly, cites its own sources, is attributed to an identifiable expert and is easy to parse will do better than vague content, whatever the markup. Structured data supports that; it doesn't replace it.

The enterprise version of the same problem

Inside companies, the same question appears as "how do we make our data understandable to AI?" The answer is the same at larger scale: shared vocabularies, stable identifiers, explicit relationships and validation. That is the work behind a governed ontology and knowledge graph, and it is what makes retrieval for AI trustworthy.

How I can help

I design ontologies and knowledge graphs for enterprises and help teams make their content and data understandable to AI systems. See ontology and knowledge graphs, or analyse your challenge and bring the result to a call.