Why do consumers delay adopting agentic search?

The same formal model of agentic commerce (arXiv 2608.08395) separates the platform's interest from the consumer's in deciding when to switch from manual to agentic search. "The platform ranks search regimes by conversion revenue, whereas the consumer also bears search expenditure." Because of this difference, "when manual inspection is sufficiently inexpensive, agentic search becomes revenue-superior before it becomes privately optimal for the consumer." The paper calls the intervening range an adoption lag: "over the intervening range, the consumer rationally continues manual search," even though the platform would already prefer she switch.

The lag arises because the platform's revenue depends only on whether a transaction completes, while the consumer's calculus also weighs the cost of search itself — time spent inspecting products or refining an agent's understanding of her preferences. While manual inspection is still cheap, a consumer keeps browsing even after agentic search would convert more revenue for the platform, because her own attention expenditure, not the platform's revenue, is what she economizes on. The paper's third finding extends this divergence to fidelity design: a platform "may provide lower AI fidelity to consumers with larger attention budgets because they can compensate through additional refinement" — an "inverted fidelity allocation" in which the consumers best equipped to tolerate a noisy agent are given the noisiest one.

This gives a mechanism for the gap the paper's own cited surveys describe — Worldpay finding only 6 percent of U.S. consumers willing to cede full purchase control to an AI assistant even though 44 percent would let one browse — by showing that rational, cost-based caution, not only distrust, can keep consumers on manual search past the point a platform wants them to switch. It sits alongside Does co-design participation hide misalignment in preference agents?: where that study finds users can over-trust a preference agent's representation of them, this model shows the opposite failure mode is also rational — under-adopting agentic search even once it would serve the platform, because the consumer alone bears the articulation cost. It is a sibling result to Why does manual search fail as product choices become complex?, from the same model.

The adoption lag and inverted fidelity allocation are derived properties of a stylized two-regime model, not measured gaps between actual platform revenue and consumer uptake; the paper offers no empirical test of the lag's size or duration in any real marketplace. The implication the model supports is that a platform pushing consumers toward agentic search before switching is privately optimal for them is asking consumers to absorb a cost — search expenditure — that the platform's own revenue metric does not register, and that observed caution about full delegation may reflect this rational calculus rather than simple distrust of AI.