By Amara Osei

PubMed climbed from 64th to fourth place among the domains ChatGPT cites most often in U.S. answers between March and September, while WebMD fell from 18th to 79th, according to Ahrefs data. Ahrefs writes that ChatGPT "seems to be prioritizing original sources over sites that summarize them."

The analysis tracked the 200 most-cited domains in U.S. ChatGPT answers, based on 2.4 million answers in March and 3.8 million in September. Most of the climbers were government health and research sites, regulators, and hands-on product testers. Britannica, Investopedia, and the New York Post dropped out of the top 200 entirely.

What does the shift mean for publisher strategy?

Ahrefs groups WebMD and Verywell Health as health sites that summarize research, and classifies Healthline and Investopedia as health and finance explainers. Both categories lost citation rank. The gainers include regulators such as the CFPB and FTC, plus support pages from Google and Microsoft, and testing sites like RTINGS.

The pattern suggests brands and publishers should focus on "pages that only you can create, such as original research, hands-on testing, and your own product information," according to Ahrefs.

One caveat: citation rank does not measure traffic. A page can be cited often and drive little referral volume. Ahrefs itself recommends checking how often ChatGPT cites your own pages for core topics — and which sources it cites instead — before acting on the trend.

How solid is the methodology?

Ahrefs' tracking system adds prompts over time, so a rank change could stem from new questions rather than a genuine change in what ChatGPT cites. To control for this, Ahrefs compared roughly 25,000 prompts that appeared among its 150,000 highest-volume prompts in both March and September. Every site on the chart moved in the same direction on that fixed prompt set, the company said. That consistency strengthens the finding, though the sample still reflects one measurement system's view of U.S. answers.

Why did Reddit and Wikipedia citations fall?

Reddit appeared in 19.1% of U.S. ChatGPT answers in March but only 2.3% in September, per Ahrefs. Wikipedia fell from 14.1% to 2.9% over the same period. Ahrefs notes both remain among ChatGPT's most-cited domains despite the decline.

The Reddit drop tracks with earlier data. Promptwatch showed reddit.com's share of ChatGPT Search citations fell from an average of 3.83% in late July and early August to 0.52% in mid-August. Ahrefs measures a different metric — the share of answers citing Reddit — so the two figures are not directly comparable. In Ahrefs' worldwide sample, reddit.com was still the ninth domain ChatGPT searched most often with site: searches, which restrict a search to a single website.

How often does ChatGPT actually search the web?

Out of 66.6 million free, logged-out GPT-5.6 Luna responses from August and September across all countries, 14.4% involved a web search. Responses that did search ran an average of around 1.6 fan-out queries — background searches ChatGPT performs before answering. OpenAI's help page states ChatGPT may search the web on its own when a question would benefit from current information.

Of the 15 million fan-outs, 6.8% were site: searches, spread across 116,000 domains. Among site: searches targeting the roughly 160 most-searched domains, 75.9% went to brand sites — company websites like apple.com or ikea.com, in Ahrefs' definition.

Other studies report much higher search volume per prompt. Nectiv found an average of 7.61 fan-outs per prompt across roughly 4,000 prompts processed through GPT-5.6 Sol via the API. Ahrefs cautions the numbers aren't directly comparable: the samples differ in models, prompts, and collection methods, and Ahrefs' 1.6 figure counts only responses that involve searching. Ahrefs also notes that free, logged-out use produces fewer searches than paid or reasoning models — a meaningful limit on how far this data generalizes.

What should marketers do next?

Ahrefs describes its fan-out data as a snapshot of a single model taken between August and September, and warns these patterns could change in future updates. For brands whose visibility depends on explainer content, the practical move is narrow: each time OpenAI updates its model, check which sources ChatGPT cites for the topics you care about.