Adobe's holiday research finds traffic arriving from AI assistants has doubled, and that this traffic engages with brand offers at a higher rate than non-AI traffic while generating more revenue per visit. Two numbers, one direction: the smallest channel in the mix is becoming the most efficient, and the efficiency is measurable enough to reallocate budget against.Adobe AI traffic study
The Two Ratios That Matter
The first ratio is engagement: AI-referred visitors interact with offers more often than visitors from other channels. That is consistent with how assistants work, since a user who asks for a recommendation has already stated a need, a budget range and often a constraint, and arrives pre-qualified in a way that keyword search rarely achieves.higher engagement from AI traffic
The second ratio is revenue per visit, which Adobe reports as higher for AI traffic. Higher engagement plus higher revenue per visit means the channel is not just better at starting sessions, it is better at finishing them, which is the harder problem and the one that justifies investment in content rather than in media.
Why Assistants Produce Better Sessions
Intent Is Stated Not Inferred
Search marketing has spent two decades inferring intent from keywords and behavior. An assistant conversation receives intent as input. The difference shows up in session depth: a visitor who has already been told which three models fit their need spends less time comparing and more time confirming.
Comparison Happens Before Arrival
Much of the comparison that used to happen on a retailer's site now happens inside the assistant. That reduces the number of pages a visitor views, which can look like weaker engagement in analytics even when the visitor is more valuable. Teams that optimize for page views will misread this shift.
Trust Transfers With the Recommendation
A recommendation from an assistant carries a small amount of transferred trust. The retailer inherits the benefit but not the control, which is why the first screen a referred visitor sees matters more than usual. If the landing page does not match the recommendation, the transferred trust evaporates in seconds.
What This Does to Channel Budgets
If AI traffic converts at a higher rate, the marginal dollar should move toward the content that assistants cite, which is not the same as the content that ranks in search. Ranking rewards keyword coverage and link authority; citation rewards structure, specificity and verifiable data.Meta AI agent for small businesses
That reallocation is uncomfortable for teams organized around search. It requires product pages written as answers, specifications expressed as comparable attributes, and claims tied to sources. It also requires accepting that some of the value will be captured by the assistant rather than by the brand's own domain, a trade-off that small merchants now face as well, since platform agents hand them a citation surface they could never build alone.Meta Muse AI agent
The Measurement Trap
Most analytics stacks attribute AI traffic to referral or direct, depending on how the assistant passes the click. Without a dedicated segment, the channel looks small and gets ignored, which is exactly the pattern Adobe's data suggests is wrong. Tagging AI referrals explicitly is the first step, and it is usually a one-week engineering task.
The second step is cohorting. Compare AI-referred customers with search-referred customers on ninety-day repeat rate, not just first-order conversion. If the assistant keeps recommending the brand, repeat rates hold. If it rotates recommendations, the first order may be the last, and the channel's true value is lower than the first-quarter numbers imply. Holiday benchmarks point the same way, with total spending projected near a trillion dollars while store traffic stays flat, which leaves conversion quality as the only variable still under a retailer's control.Holiday season outlook
Where Retailers Should Start
Start with the ten product questions customers ask most often, and rewrite the answers so they can be extracted without context. Then add structured attributes that assistants can compare, and publish the sources behind any performance claim. This is unglamorous work, and it is the only durable way to be cited consistently.
Then measure what changed. Citation share in the target question set, assisted conversion, and repeat rate by channel give a complete picture within one quarter. Retailers that build this baseline now will be able to argue for budget with evidence when the channel becomes large enough to matter to the whole organization.
