Should GEO Work Start With Schema Markup?

Schema markup is often one of the first things people touch when starting GEO (generative engine optimization) work. "Add JSON-LD schema markup if you want to be cited by AI" is advice you hear often. Whether it actually translates into GEO results is something Ahrefs tested directly.

Ahrefs' Greg Secrist selected 1,885 pages that newly added schema markup and compared their citation changes, before and after, against similar pages that didn't add schema. Rather than simply comparing pages with schema to pages without it, the study tracked each page's own before-and-after change — a design that gets closer to causation than mere correlation. Results were measured across Google AI Overviews, Google AI Mode, and ChatGPT.

The results were clear. Google AI Overviews citations actually dropped 4.6%, while AI Mode and ChatGPT saw increases of 2.2% and 2.4% respectively — gains too small to call statistically meaningful. Pages with schema markup did tend to be cited more often (a correlation), but adding schema didn't actually cause citations to rise (no causal effect). Ahrefs concluded that "schema alone didn't move the needle on AI citations."

That doesn't mean schema markup is useless — it's still needed for traditional search engines' rich results (star ratings, pricing, and other extras shown in search listings). But putting it first in line for GEO budget or time isn't well supported by this data. Ahrefs found that content quality itself has a far bigger effect on whether something gets cited.

In Korea

Schema markup is sometimes treated as the first step of GEO here too, but this experiment suggests it mainly dresses up how a page appears in traditional search results, not whether AI cites it — not really a GEO lever at all. If resources are limited, checking the specificity and credibility of the content itself first is the better move from a GEO standpoint.