SimilarWeb: AI-Recommended Brands Get 2.5x More Site Visits, Mostly via Branded Search
A SimilarWeb study published on 21 June 2026 measured something most AI-search research has skipped: what happens after a chatbot recommends a brand. The headline finding is that being recommended in ChatGPT made a brand 2.5 times more likely to receive a site visit within the following seven days, and that most of that traffic arrived not as a visible AI referral but as a branded Google search.
Most existing data on AI search measures the downside, the click-through and zero-click losses from AI Overviews answering queries in place. This study, “The Downstream Impact of AI Visibility”, measures the upside and, more usefully, the mechanism by which AI influence reaches a site.
What the study found
SimilarWeb used panel-based clickstream data to track US desktop users who received a ChatGPT recommendation, then watched where they went next. Three verticals were covered, finance, travel, and beauty, using head-to-head brand pairs such as American Express against Capital One, Skyscanner against Kayak, and Sephora against Ulta.
- 2.5x more site visits. Users who saw a brand recommended in ChatGPT were 2.5 times more likely to visit that brand’s site within seven days than users who did not.
- 55.9% arrived via branded search. Most of that downstream traffic reached the site through a branded query in traditional search, not through a click on a citation inside ChatGPT.
- Deeper engagement. AI-influenced visitors averaged 12.0 pages and 11.8 minutes per session, against 6.5 pages and 5.6 minutes for other visitors.
Why the branded-search figure is the real story
The 55.9% number explains a measurement problem this site has covered repeatedly. When someone asks ChatGPT for a recommendation, sees a brand, and later searches that brand name on Google, the visit lands in analytics as branded organic search. There is no AI referral in the path. The AI did the persuading; Google got the attribution.
This is why measuring AI’s impact by referral traffic alone undercounts it. Anyone judging the value of AI visibility by the size of the “ChatGPT” or “Perplexity” referral line in GA4 is reading a fraction of the effect. The rest is hiding in branded search, and some of it in dark traffic that carries no referrer at all.
It also supports a position this site already holds: the click-free visit model is real, but its magnitude is unsettled. This study puts a number on the mechanism without resolving the magnitude, because the panel is narrow.
What this means
The practical takeaway is about attribution, not tactics. If AI recommendations convert into branded search, then branded search volume becomes a partial proxy for AI visibility, and a sudden rise in branded queries may reflect AI exposure rather than brand-marketing spend. Pairing branded-search trends with direct AI-visibility tracking gives a fuller picture than either alone.
The caveats matter and should temper any firm conclusion. The study covers US desktop only, explicitly excluding mobile and international traffic, and rests on three verticals and a set of named brand pairs rather than a whole-web sample. SimilarWeb states the relationship is a correlation, not proven causation, and notes plans to extend the research to mobile and more categories. Rand Fishkin contributed context from earlier SparkToro work showing that AI recommendations vary between repeated queries, a reminder that “being recommended” is not a fixed state.
Treated as directional evidence rather than a settled measurement, the study is a useful corrective: the payoff from AI visibility shows up as branded search and deeper on-site engagement, not as a tidy AI-referral line in your analytics.
Sources
- The Downstream Impact of AI Visibility — SimilarWeb
- Being Recommended in AI Makes Users 2.5x More Likely to Visit Your Site — SimilarWeb
- AI-Recommended Brands Saw 2.5x More Site Visits: SimilarWeb — Search Engine Journal
- ChatGPT recommendations drive more brand website visits: Study — Search Engine Land
- Your analytics are lying: SimilarWeb traces AI recommendations to real traffic — PPC Land
More news
-
Bruce Clay, a Founding Figure of SEO, Has Died
Bruce Clay, widely credited with first using the term search engine optimisation and founding one of the earliest SEO agencies, has died.
-
Google Cloud Publishes the Open Knowledge Format for AI Agents
OKF is a draft open spec for representing curated knowledge as Markdown files with YAML frontmatter. Google scopes it to internal data, but SEOs are asking whether it could work for websites.
-
Cloudflare and Browser Makers Announce PACT Bot-Verification Protocol
Cloudflare, Google, Microsoft, Mozilla and Shopify are developing PACT, a privacy-first token protocol to separate humans and authorised agents from abusive bots.