SEO tool provider Ahrefs has published changes in the rankings of domains cited in ChatGPT’s U.S. answers. Comparing March with September, the medical research database PubMed rose from 64th to fourth place, while health information site WebMD fell from 18th to 79th. Sites with original information, such as research data and hands-on product tests, have moved up the rankings.
Research databases surge as explainer sites fall back
The analysis covered the top 200 domains most frequently cited in ChatGPT’s U.S. answers. It was based on about 2.4 million answers in March and about 3.8 million in September.
The sites that climbed the rankings were mainly PubMed, government health and research agencies, regulators, and testing sites that evaluate products firsthand. Examples cited include the FDA and RTINGS.
By contrast, explainer sites that break down research for readers, such as WebMD and Verywell Health, as well as health and finance sites like Healthline and Investopedia, fell in the rankings. Britannica, Investopedia, and the New York Post dropped out of the top 200.
Ahrefs says ChatGPT appears to favor original sources over sites that summarize information.
An era of choosing ‘owners of information’ over ‘summarizers’?
Health and finance explainers have traditionally been strong performers in search engines. They have helped readers understand research papers and public documents by recasting them in plain language. But when AI generates an answer, it can take on that very process of “translation” itself. If so, it seems natural that AI would turn to the original source rather than the person who translated it.
Another striking development is the rise of product-testing sites. Unlike articles that simply rearrange spec sheets, these sites offer measured results and user experience that can only be found there. The criteria for being cited by AI may be coming down to whether a site has original content.
Ahrefs also recommends that site operators focus on pages they alone can create. Examples include original research, hands-on product tests, and information about their own products.
Share of citations for Reddit and Wikipedia drops sharply
The share of U.S. answers citing Reddit fell from 19.1% in March to 2.3% in September. Wikipedia also dropped, from 14.1% to 2.9%. Ahrefs says both remain among the domains ChatGPT cites frequently.
Another analytics tool, Promptwatch, reported a similar decline for Reddit in August. The share of ChatGPT Search citations going to reddit.com reportedly fell from an average of 3.83% in late July and early August to 0.52% in mid-August. However, Ahrefs measures the “share of answers that cited Reddit,” so its methodology differs from Promptwatch’s.
In a separate global sample, reddit.com ranked ninth among the domains ChatGPT most often specified in site: searches (searches limited to a particular site). This suggests Reddit still has a presence in search even after its share of citations fell.
Citation rankings and traffic are different things
It is important to note that these rankings measure how often a site is cited, not how many visitors it receives. A site’s traffic cannot be assumed to have fallen by the same amount just because its ranking dropped. Conversely, ranking near the top does not necessarily mean people are clicking through and reading the site.
Also, Ahrefs’ tracking system adds questions to its analysis over time, so changes in the questions included could have affected the rankings. To account for this, the company compared about 25,000 questions that ranked among the top 150,000 high-volume questions in both March and September. It says every site shown in the chart moved in the same direction.
The reality: most answers don’t involve a search
Another analysis looked at how often ChatGPT searches the web. Of 66.6 million answers given to free, logged-out users in August and September, 14.4% involved a web search. The analysis covered the GPT-5.6 Luna model and data from all regions.
For answers that involved a search, an average of about 1.6 searches—known as fan-outs—were run in the background before the answer was generated. Of 15 million fan-out searches, 6.8% were site: searches limited to particular sites, covering 116,000 domains. Among the roughly 160 most-searched domains, 75.9% of searches were directed at companies’ official sites, such as apple and ikea.
Another research firm, Nectiv, processed about 4,000 questions through the API using GPT-5.6 Sol and found an average of 7.61 fan-outs per question. However, the figures cannot be directly compared with 1.6 because the model, questions, and collection methods differed. Ahrefs also says one factor may be that free, logged-out users trigger fewer searches than paid models or reasoning models.
What site operators should verify
Ahrefs’ findings are a snapshot of a single model, based on data collected from August to September. The company also notes that the trend could change with future model updates.
For site operators, a practical step is to regularly check which of their pages ChatGPT cites on their core topics—and which sources it chooses instead. If explanatory articles are their main source of traffic, it is worth revisiting this whenever OpenAI updates its models.
Comments