Brands recommended by ChatGPT are 2.5 times more likely to receive a website visit within seven days than competing brands that were not recommended, according to a new study by Similarweb covering US desktop behavior from July through December 2025.
The finding is not primarily a referral story. Similarweb tracked users who asked ChatGPT industry-relevant questions, received a named brand recommendation, and then visited a brand website within the following week. Only 19.9% of those AI-influenced visits arrived via direct navigation. The majority, 55.9%, arrived through search, compared with 40.4% of search visits in the non-AI-influenced control group. The practical implication for measurement teams: a ChatGPT recommendation may be driving branded query volume that shows up in Google Search Console as organic search, with no AI referral tag in sight.
The data held across three verticals. In finance, a ChatGPT recommendation for American Express sent 7.2% of influenced users to AmEx within seven days, against 3.1% going to Capital One. The reverse pattern was equally clear: after a Capital One recommendation, 14.2% visited Capital One versus 3.8% for AmEx. In travel, Kayak recommendations pushed 12% of users to Kayak and just 3.4% to Skyscanner. In beauty, Sephora and Ulta each showed similar asymmetry when recommended.
Similarweb excluded users who had visited the brand’s site in the prior four weeks or who had named the brand in their original prompt, which filters out people who were already in-market for that specific brand. The 2.5x lift therefore reflects incremental discovery, not confirmation of existing intent.
Engagement depth widened the gap further. AI-influenced visitors viewed an average of 12 pages and spent 11.8 minutes on site. Non-AI-influenced visitors averaged 6.5 pages and 5.6 minutes. Similarweb attributed the difference to decision-narrowing that happens inside the AI conversation before the user ever reaches the brand site. By the time they click, they are pre-qualified.
The branded-search attribution angle is the piece most attribution models will miss. Standard last-click or session-based attribution assigns these visits to organic or branded paid search, not to ChatGPT. A GEO team trying to justify AI visibility investment will not find the signal in referral reports. The cleaner measurement approach is to track branded query volume in Search Console before and after periods when a brand gains or loses AI recommendation share, then compare against competitor branded query trends.
Search Engine Land reported the study findings on June 24, 2026, citing Similarweb’s full report on the downstream impact of AI visibility.
GEO teams that are not yet tracking AI recommendation share by query cluster are flying blind on a channel that is already moving branded search volume in finance, travel, and beauty. The next step is instrumenting branded query monitoring now, before a competitor recommendation surge makes the gap visible in quarterly reporting.
Search Engine Land reported on the Similarweb study covering US desktop user behavior from July through December 2025, published June 24, 2026.
