Most of the conversation around AI and ecommerce has focused on volume, how much traffic generative AI tools are sending to retail sites, and how fast that number is growing.
That framing misses a more useful question. Once someone actually arrives at a retail site through an AI assistant, how do they behave once they're there, and does that behaviour tell retailers anything about how ready these visitors are to buy?
Adobe's own analytics data, drawn from the retail sites it tracks, gives a clearer answer than most of the growth headlines suggest. Comparing generative AI referral traffic to all other traffic sources in February 2025, Adobe found that AI-referred visitors viewed 12% more pages per visit and showed a 23% lower bounce rate than shoppers arriving through search, social, or direct channels.
That combination of higher page depth and lower bounce rate points to something worth paying attention to, these are not casual, low-intent clicks. They're visitors staying on site, exploring further, and behaving in ways that typically signal genuine interest rather than an accidental landing. Searchflex has flagged this exact pattern as one of the more telling details in the dataset.
The Contradiction Worth Explaining
Here's where the data gets genuinely interesting. Despite browsing more and bouncing less, AI-referred visitors in February 2025 were still 9% less likely to convert than traffic from other sources. That's a real gap, even though it had narrowed sharply from July 2024, when AI traffic converted 43% worse than everything else. So the pattern isn't simple. These are engaged visitors who explore more of a site, yet still purchase less often than shoppers arriving through other channels. That combination doesn't fit the usual assumption that engagement and conversion move together.
Why Engagement And Conversion Might Be Decoupled Here
The most plausible explanation sits in where these visitors are in their decision-making process. Someone typing a question into an AI assistant is often doing research rather than executing a decision they've already made. Compare that to someone searching a specific product name or clicking a retargeting ad, both of which usually carry strong navigational or transactional intent baked in. Conventional search traffic includes enormous volumes of people who already know exactly what they want and are simply looking for the fastest route to buy it. AI-referred traffic, by contrast, may be arriving earlier in the comparison and consideration phase, browsing multiple options thoroughly precisely because a decision hasn't been made yet.
This doesn't mean the engagement is meaningless. Extended browsing and repeated page views on a single site suggest the visitor is seriously weighing that retailer against alternatives, not simply passing through. It's a different kind of visit to the transactional, high-intent click that conventional search is often optimised around.
What This Means Beyond A Single Statistic
It would be a mistake to treat this as proof that AI traffic definitely converts later rather than not at all. The data doesn't establish causation, and retailers should be cautious about drawing firm conclusions from a single behavioural pattern measured over one month. What it does suggest is that AI-referred visitors deserve a different measurement lens than the one typically applied to search or social traffic. Judging this channel purely on immediate conversion rate risks undervaluing visitors who are still in the process of deciding, and who may convert later through a different channel entirely once their research concludes.
The Practical Takeaway For Retailers
For ecommerce brands trying to make sense of a new and rapidly growing traffic source, the behavioural data matters more than the growth percentage alone. A channel bringing in visitors who browse deeply and stick around, even if they're not converting immediately, is a channel worth understanding properly rather than dismissing on conversion rate alone. Retailers that track AI referral behaviour over a longer window, rather than judging it against the same immediate-conversion benchmark used for paid search, are more likely to spot where this traffic actually contributes to revenue, even when that contribution doesn't show up in the first visit.
That distinction matters more as this traffic source keeps growing. A channel that behaves differently to search or social needs to be measured differently too, and retailers who apply old benchmarks to new behaviour risk making decisions based on an incomplete picture. Watching how the conversion gap continues to close in the months ahead will likely say more about where this channel is heading than any single month's snapshot can.