AI agents now read shopping sites constantly but barely click through — what's really happening in that silent gap?

This explores why AI agents and crawlers now read retail and travel sites heavily while sending back very little referral traffic, and which measures besides clicks help explain the gap. A caveat first: none of the retrieved notes measures crawler volume against referral volume. The corpus can't give you the ratio itself. What it does offer is a set of explanations for why that ratio may be the wrong thing to watch. Read together, they suggest the gap isn't a measurement error. It's what happens when the agent, not the person, does the comparison shopping.

The simplest explanation is that the agent finishes the job the click used to start. Do AI shopping agents destroy search ad revenue while leaving influencer marketing intact? argues that shopping agents squeeze discovery and purchase into one zero-click step. The agent reads the pages, but nobody visits them. The same note draws a useful line between meeting a need, which agents absorb, and creating a want, which they can't. That suggests referral traffic will fall hardest for need-driven searches, while influencer and social channels hold up. So one metric worth tracking is the mix of intent types, not just the total number of visits.

A second explanation concerns who keeps the demand signal, meaning the information about what the shopper wanted. Are AI agents making attention-based ads irrelevant for aggregators? argues that when an agent ranks options against the shopper's stated criteria, the aggregator never sees the intent it used to sell ads against. The note's concrete figure is Tripadvisor's Hotels segment revenue, which fell 23% in Q2 2026, with management blaming structural shifts in search traffic. Here the useful measure is revenue per unit of intent the business gets to see, not crawl volume. Heavy crawling with falling segment revenue is exactly the pattern you'd expect. Will agents compete for attention just like users do? states the general version: the thing businesses compete for shifts from clicks to being chosen by the agent. That calls for new metrics, such as how often an agent picks you, which current referral analytics don't capture.

The less obvious angle is that platforms and shoppers may score agentic search differently, so the gap can last. Why do consumers delay adopting agentic search? models a platform that judges agentic search purely by how much revenue it converts, while shoppers also count their own search effort. The result is a rational lag in which people keep searching by hand even after switching would earn the platform more. One reading is that heavy crawling reflects platforms getting ready ahead of actual consumer use. That would make the low referral numbers partly a timing gap rather than a failure. Incentives matter too. Can an AI agent serve both merchant and user interests fairly? argues that agents paid by merchant referral fees can't serve both sides loyally. Whether a referral happens at all can depend on the business model, not on how relevant the merchant is.

The takeaway you might not have expected: the field may be measuring the wrong thing. The metrics that matter seem to be how often agents choose you, the mix of need-driven versus want-driven intent, and conversion revenue per query. Referral clicks reflect a funnel that agents are taking apart. For the actual crawler-to-referral numbers, you'll need sources beyond these notes.