Does structured data help AI engines cite you?
The question every SEO asks about GEO: does my schema markup matter to ChatGPT? The honest answer is less direct than vendors claim — but the underlying instinct is right.
What the evidence supports
AI engines read pages the way a headless browser does — rendered HTML, main content, links. What helps:
- Clean, extractable content. Pages where the answer sits in the HTML, not behind JS rendering or interstitials. If a crawler can’t read it, the fan-out can’t use it.
- Clear structure. Headings that restate the question, lists and tables — the model lifts chunks, and well-chunked pages lift cleanly.
- Direct claims. “X costs $49/mo” is citable; “competitive pricing” isn’t. The model needs a sentence it can attribute.
Schema.org helps indirectly: the same markup that earns rich results makes your page legible to retrieval systems. Article, FAQPage, Product markup clarifies what’s on the page. It’s a readability signal, not a citation button.
What doesn’t matter for citations
- Submitting to “AI crawlers” lists
- Blocking or allowing GPTBot differently — retrieval for answers happens at search time, not training time
- AI-specific meta tags — none exist that engines honor
The measurable version
Don’t theorize — observe. Run your prompts through the monitor API and look at which of your pages appear in sources[]. Check what format they use. The pages that get cited share traits you can replicate; the ones that don’t usually fail on extractability, not schema. How engines choose citations covers the full selection path.