Schema markup is hygiene, not a growth lever
Structured data earns rich results and entity clarity. The evidence that it lifts AI citations on its own is thin.
Adding JSON-LD is one of the most confidently recommended “GEO tactics” going, and one of the least supported by evidence.
Controlled comparisons, with Ahrefs’ study across roughly 1,885 pages being the widely cited one, found no meaningful lift in AI citations from adding structured data alone. Live fetch tests point the same direction: engines read the visible HTML. Hidden script tags aren’t where the answer comes from.
So why ship it anyway
We put Organization schema on this site, and our site audit flags it when your homepage has no JSON-LD naming your organization or product. Both for defensible reasons:
- It clarifies entity identity for search consoles and rich results
- It’s cheap, standardised, and part of a professional baseline
- It costs an afternoon, so there’s no good reason to skip it
None of which is the same as “add schema and get cited.” We try to be careful about that distinction, because the industry mostly isn’t.
What the research points at instead
The Princeton GEO work (Aggarwal et al., KDD 2024) found that adding statistics, quotations, and citations to visible copy raised position-adjusted visibility meaningfully in their benchmark. Keyword stuffing made things worse.
That maps onto something intuitive: a model looking for a quotable claim will take the page that contains one.
What makes a page worth citing
- A direct answer near the top of the page, not four paragraphs down
- Question-shaped headings
- At least one figure, or an external source worth quoting
- A visible FAQ or comparison section
- Evidence the page is maintained, such as a visible updated date or
dateModified
The best benchmark is the page that’s already being cited for your question. Open it next to yours and compare. Our prompt results list those sources for every tracked question.
Schema is one line on that list. Visible, quotable substance is the work.