Home » News » Getting Recommended by ChatGPT Sends 2.5x More Traffic Your Way June 25, 2026 3 mins read Getting Recommended by ChatGPT Sends 2.5x More Traffic Your Way By Daniel Beauregard The latest research by Similarweb has offered potentially the first ever empirical evidence that AI recommendations can actually impact website traffic. As per the findings of the study titled, The Downstream Impact of AI Visibility, published by Similarweb, brands appearing in ChatGPT recommendations were 2.5x more likely to receive a site visit within seven days than their peers which were not part of recommendations. These findings reported by Matt G. Southern at Search Engine Journal will certainly alter the way marketers look at AI visibility, branded search, and attribution. How the Study Was Conducted This is not about modeling estimates or survey recalls but clickstream data from Similarweb’s opted-in US Desktop web panel. In the period from July 2025 to December 2025 and January 2026 survey, Similarweb conducted the research on three industry verticals – Finance, Travel, and Beauty. The three brand pairs selected in the research were – American Express & Capital One in Finance, Skyscanner & Kayak in Travel and Sephora & Ulta in Beauty as they compete in the same space. The study design was very specific where the researchers studied the clickstream of users who submitted their queries to ChatGPT and received recommendations from AI on particular brand(s), to see whether they visited either of those websites or their competitors’ sites in the next seven days after their interactions. They excluded users who visited any of the brand’s sites in previous four weeks or mentioned them in their prompt. What the Numbers Show The main takeaway is the 2.5x uplift of site visits within seven days of an AI recommendation compared to the non-recommended competitor brands. However, there are more data points in the report. The uplift occurred in all three categories. In the Finance category, 7.2% of users visited American Express after receiving a recommendation while only 3.1% visited Capital One. Also, Similarweb noted the high level of engagement of users who visited these sites after a recommendation. They visited, on average, 12 pages and stayed for 11.8 minutes, in contrast to those who arrived to the sites via other sources – they viewed 6.5 pages and stayed on the websites for 5.6 minutes. According to Similarweb, such users had probably already narrowed down their options during an AI conversation. The Traffic Shows Up in Search, Not in Referrals Another interesting finding of the Similarweb report is where exactly this traffic comes from. The majority of it comes neither from AI referral nor referral traffic at all. 55.9% of AI-influenced visits went through search engines, compared with 40.4% of non-AI-influenced visits. Also, 19.9% of AI-influenced visits came directly from the user, whereas the percentage of direct traffic for the standard visits was 38