Ahrefs (2026) — schema markup and AI citations

Summary:

This piece is about a simple SEO question: if you add schema markup to pages that already get cited by AI tools, do those pages get cited more.

Ahrefs published it. Ahrefs sells SEO and AI visibility tools, including products used in the study, so it has a clear commercial angle. The goal was to test cause, not just correlation, because pages with schema often also have better SEO in general.

Ahrefs says it first looked at 6 million URLs and saw that AI-cited pages were much more likely to have schema. But the main study is the stronger part.

The team tracked 1,885 pages that added JSON-LD schema and matched them with 4,000 similar control pages that did not add it.

They used crawl history to find when schema first appeared, then compared AI citations in the 30 days before and after that date across Google AI Overviews, Google AI Mode, and ChatGPT.

They also checked the result four ways, including a matched difference-in-differences analysis, simple group comparisons, week-by-week trend checks, and a repeat analysis with a different time window.

The main finding is that adding schema did not meaningfully lift citations on any platform they tested.

In their preferred analysis, Google AI Mode was up 2.4% and ChatGPT was up 2.2%, but both were small enough to look like noise.

Google AI Overviews showed a 4.6% relative decline for treated pages, but Ahrefs says that drop was small, happened while both treated and control pages were already falling, and cannot be cleanly blamed on schema alone.

Across all four tests, the big story stayed the same: no clear positive effect.

The article also says 53% of AI-cited pages have schema, but argues this likely reflects stronger sites doing many good SEO things at once, not proof that schema itself drives citations. It also notes that pages in the sample were already heavily cited before schema was added, so the study does not answer whether schema helps pages get discovered in the first place.

There are several notes.

This is not a randomized experiment, so other page changes may have happened at the same time as schema.

Ahrefs pooled many schema types together, only studied JSON-LD in HTML, and only used a 30-day post-change window. The sample was limited to pages already getting lots of AI Overview citations, so the findings may not carry over to low-visibility pages. Ahrefs also used its own crawler data and its own products in the work, which raises vendor-interest concerns. Still, the methodology is described clearly.