Original insights by Laurie Naspe, Sam Sheridan, and Adelle Kehoe
30-second rundown
Key learnings:
- AI recommendations move people: Similarweb found that users were 2.5 times more likely to visit a ChatGPT-recommended brand within seven days.
- Analytics misses the first influence: 55.9% of AI-influenced visits arrived through branded search, so search received credit for demand that began in ChatGPT.
- Prominence must meet protection: Track AI recommendations alongside branded-search demand and make sure your official site is easy to find when people search your name.
This week: Choose five buyer questions that could trigger a brand recommendation, record whether ChatGPT names you or a competitor, review the branded searches and direct visits reaching your site, and check whether your official pages dominate the results for your name. Save the results as a simple baseline for future comparison.
Coming 3 months: Set up an AI search monitoring tool to repeat selected prompts daily and combine the results with branded-search clicks, direct visits, and optional customer survey answers. Assign one analytics owner to review the combined pattern monthly and investigate only changes that persist across several measures.
Laurie Naspe, Sam Sheridan, and Adelle Kehoe followed the footprints that ordinary analytics abandons in the desert. Their Similarweb research asks the question haunting every marketer with an AI visibility chart: when ChatGPT recommends a brand and nobody clicks the answer, does anything happen afterward?
It does. The visitor simply returns by another road, while the dashboard hands Google the medal and pretends ChatGPT had nothing to do with it.
1. The Recommendation Followed the User Out of the Chat
Similarweb tracked thousands of real user journeys on United States desktop devices across finance, travel, and the beauty industry. The researchers followed people who asked ChatGPT an industry-related question, received a specific brand recommendation, and then browsed the web during the following seven days.
They excluded people who had named the brand in their prompt or visited its website during the previous four weeks. That matters because the study was trying to observe new influence, not catch an existing customer already circling the building with cash in hand.
Recommended brands were 2.5 times more likely to receive a visit than competing brands that were not recommended. The pattern appeared across all three industries, and when ChatGPT recommended the competitor, the later traffic followed the competitor instead. The mention was not decorative confetti; it changed which door people approached.
2. Branded Search Wore the Stolen Uniform
The most useful finding is not merely that people visited. It is how they arrived. A person could read the recommendation on Monday, remember the name on Thursday, search for it, and appear in the analytics report as an ordinary search visitor with no visible connection to ChatGPT.
55.9% of AI-influenced traffic arrived through search, compared with about 40% of standard visits. In plain terms, ChatGPT often creates brand demand without sending a trackable referral, then Google collects the credit like a crooked tollbooth operator on an empty highway.
This makes branded search results part of the AI customer journey. If a comparison page, reseller, or competitor advertisement controls the results for your name, the recommendation can still leak value before the visitor reaches you. Being named by the oracle is only half the job; owning the road back to your site is the other half.
3. The Visitors Arrived Hungry, but the Study Did Not Watch Them Buy
AI-influenced visitors viewed an average of 12.0 pages and spent 11.8 minutes on site. Other visitors averaged 6.5 pages and 5.6 minutes, so engagement was roughly twice as deep after an AI recommendation. These people did not stumble in, lick the wallpaper, and flee.
That is a strong commercial signal, but not a proven sale. The study observed website behavior, not completed purchases, and it covered United States desktop activity in three large consumer categories. The findings may not transfer perfectly to mobile users, smaller brands, business-to-business buying, or a six-month sales cycle involving fourteen nervous executives.
The honest response is measurement, not worship. Compare AI recommendations, branded-search demand, direct visits, and customer self-reporting over time. When several independent signals rise together, you have something sturdier than a referral report and less deranged than assigning revenue to every chatbot mention.
AI visibility is no longer just a vanity metric waving a tiny flag in a dashboard. Similarweb has shown that a recommendation can escape the chat, haunt the buyer for days, and reappear as search traffic wearing a false moustache.
Measure the whole journey and protect the branded road home. Otherwise the oracle whispers your name, the customer moves, and your analytics applauds the wrong machine.
