Pages that added JSON-LD schema markup did not earn more citations from AI platforms as a result. Ahrefs published a controlled study on May 12, 2026, tracking 1,885 web pages that introduced JSON-LD between August 2025 and March 2026, finding no meaningful citation uplift across Google AI Overviews, Google AI Mode, or ChatGPT.
Study Design and Methodology
Ahrefs tracked 1,885 web pages that added JSON-LD schema between August 2025 and March 2026, matched them against 4,000 control pages, and measured citation changes across Google AI Overviews, AI Mode, and ChatGPT. Using its Brand Radar tool and Agent A, Ahrefs conducted a matched difference-in-differences analysis to account for platform trends.
The methodology was designed to separate the effect of schema from the broader signals that tend to accompany it. Schema markup tends to live on better-maintained, more technically sophisticated sites, and those same sites publish stronger content, build more authority, earn more links, and do all the other things that get pages cited. For each treated URL, the researchers picked three control URLs from different domains, with similar pre-period citation levels, that had never added JSON-LD.
Citation rates were measured for each page across Google AI Overviews, Google AI Mode, and ChatGPT in the 30 days before and 30 days after each page's schema addition date. Four separate analytical tests were run. The difference-in-differences test, described in the report as the most reliable of the four, produced the headline results cited in this article.
Results Across All Three Platforms
Google AI Overviews showed a −4.6% change relative to controls, described as a small but statistically notable decline. Google AI Mode showed +2.4%, and ChatGPT showed +2.2%, both too small to distinguish from random variation. Three additional tests were run alongside the primary comparison, and all four found no clear positive or negative effect.
The AI Overviews Decline in Context
The −4.6% result for Google AI Overviews was statistically significant, but the study's authors advise caution in interpreting it as a schema-caused harm. Both treated and control pages were already declining before the schema was added. Treated pages declined slightly faster, but the difference amounts to about 12 fewer daily citations per page in a sample where most pages were receiving hundreds. The report notes that the decline could reflect a small negative effect from schema, or it could be a coincidence, and draws no firm conclusion either way.
Who Was Studied, and Who Was Not
The dataset carries a significant scope limitation. Every page studied was already being cited heavily by AI. Every page in the dataset had 100 or more AI Overview citations in February 2025, before any schema was added, meaning these pages were already inside the consideration set, being crawled and surfaced by large language models.
If a page is already being picked up, the data suggests that adding schema is not going to push it higher. For pages that are not yet being seen by AI systems at all, schema markup might still play a role in helping them get crawled, parsed, or indexed in the first place. The study does not address that scenario.
The report also acknowledges additional methodological constraints. Pages adding JSON-LD often change other page elements at the same time, making it difficult to fully isolate the schema as a variable. All schema types were pooled together, so some individual types may behave differently. And the 30-day measurement window may be too short to capture any slower-acting effects.
What AI Systems Actually Read During Direct Retrieval
A separate experiment by searchVIU, cited in the Ahrefs report, bears directly on why schema may not affect citations at the point of AI retrieval. In tests conducted in October 2025 to determine whether AI systems use schema markup during direct page fetch, the results showed that current AI chatbots do not use JSON-LD schema markup during direct retrieval; they exclusively extract visible HTML content. Hidden structured data is ignored entirely. Neither hidden Microdata nor hidden RDFa were recognized by any system tested.
The searchVIU report notes a key limitation of this finding: schema markup could still be used during the indexing phase, in LLM training data, or in search engine-integrated AI systems such as Google AI Overviews and Bing Copilot. The direct-fetch result does not foreclose a role for schema earlier in the pipeline.
The Correlation That Preceded the Study
The controlled experiment was prompted by a correlational finding that had circulated widely in the SEO community. An initial analysis of 6 million URLs found that pages cited by AI were roughly three times more likely to include JSON-LD, a gap that had been treated by many as evidence that schema improves AI visibility.
The Ahrefs data shows a correlation: pages with schema are cited more often by AI, but the report interprets this as a sign of overall site quality rather than schema's direct impact. This interpretation is consistent with prior research. A December 2024 study from Search Atlas found no correlation between schema markup coverage and citation rates, with sites carrying comprehensive schema not consistently outperforming sites with minimal or no schema markup.
Practical Implications for SEO and Content Professionals
For pages that are already receiving AI citations, the Ahrefs data does not support using JSON-LD addition as a tactic to increase citation frequency. Schema continues to serve documented functions outside AI citation, including rich result eligibility, Knowledge Graph entity recognition, and structured data for voice and knowledge-based surfaces, and those use cases are unaffected by this finding. The study's scope means its conclusions apply specifically to pages already visible to AI systems; the question of whether schema aids discoverability for pages with no existing AI presence remains open and untested by this data.
The Ahrefs report concluded that while schema supports rich results and knowledge graphs, adding JSON-LD does not increase AI citations for pages already cited. The full study, including methodology, charts, and all four analytical tests, was published on the Ahrefs blog on May 12, 2026.
