This website uses cookies

Read our Privacy policy and Terms of use for more information.

The Maze: Similarweb connected ChatGPT brand recommendations to later website visits. The useful part is not the AI referral click. It is the delayed branded search that happens after the answer has already shaped intent. That makes AI visibility less like fluffy awareness and more like a leaky attribution problem for commerce teams.

  • AI is creating demand that analytics may hand to search. The study tracked US desktop users from July to December 2025 after they received a ChatGPT brand recommendation. Brands that appeared in those answers were 2.5x more likely to get a site visit within seven days than competing brands that did not. The study excluded users who already named the brand in their prompt, so the measured behavior is closer to assisted discovery than simple brand recall.

  • The channel mix is the awkward bit. Similarweb found that 55.9% of AI-influenced visits came through search, versus 40.4% for non-AI-influenced visits. Direct traffic moved the other way: 19.9% for AI-influenced visits versus 38.8% for non-influenced visits. AI referral traffic itself was only 8.8%. Translation: the AI answer may create the intent, but the analytics dashboard often gives the credit to branded search.

  • The users also looked more serious once they arrived. AI-influenced visitors generated 12.0 pages per visit and spent 11.8 minutes on site, compared with 6.5 pages and 5.6 minutes for non-AI-influenced visitors. That does not prove revenue or conversion. It does suggest the recommendation is not just trivia. It can move users into deeper consideration, then disappear from the visible attribution chain.

  • This is why referrer buckets are necessary but not enough. Google Analytics can be configured to group traffic from ChatGPT, Gemini, Copilot, Claude, Perplexity, and similar tools into custom AI chatbot channels. That captures sessions with recognizable AI referrers. It does not catch the user who asks ChatGPT, remembers Sephora, and searches the brand two days later.

  • The operating question shifts from traffic source to recommendation share. Similarweb already sells AI visibility tooling that tracks where brands appear across ChatGPT, Perplexity, Gemini, Grok, and Copilot. The downstream study adds the sharper commercial question: which recommendations create later visits, and which competitors capture the branded search that follows.

Why it matters: Retailers, marketplaces, travel brands, finance players, and beauty sellers will not manage AI discovery by staring at AI referral lines. The budget fight moves upstream. Teams need to know whether they appear in AI answers, whether those answers create branded demand, and whether paid search, SEO, and analytics are crediting the wrong part of the journey.

Reply

Avatar

or to participate

Keep Reading