Ahrefs Studies ChatGPT Query Fan-Out in 66.6 Million Responses

Ahrefs has published research into how ChatGPT retrieves information, based on an analysis of 66.6 million responses and 15 million associated fan-out queries. The study focuses on what that retrieval process could mean for visibility in AI-generated answers, a topic of interest to marketers assessing how audiences find information.

The responses examined came from free, logged-out use of GPT-5.6 Luna. Ahrefs says the material covered August and September. Those details define the setting described in the study summary; the summary does not establish whether the same results would apply to other models, account types or periods.

The summary identifies the scale of the analysis but provides limited information about its findings. It mentions a figure of 14.4% of the 66.6 million responses, without completing the description of what that percentage measures. It would therefore be misleading to interpret that figure as a rate of searching, citation or any other specific behavior.

Likewise, the two headline totals should be read separately: one describes responses, while the other counts fan-out queries. The summary does not explain how the queries were distributed among responses, which sources appeared in the results or how the researchers assessed visibility. It also does not provide enough detail to draw conclusions about which marketing tactics performed best.

The distinction matters when evaluating research on AI retrieval. A large set of responses can provide a substantial basis for analysis, but the size alone does not explain the method or establish a recommendation. Readers seeking the study’s conclusions would need to review the full research before assigning meaning to the incomplete percentage or applying its results to a particular campaign.

What this means for marketers

Use the Ahrefs study as a starting point for examining how AI systems retrieve information, not as evidence that a specific optimization tactic will increase visibility. Before changing a content or measurement strategy, check the full findings, the definitions behind the reported figures and the scope of the data. Pay particular attention to whether the conditions studied match the audience or AI product relevant to your work.