AI shopping agents and the reallocation of consumer attention: seller visibility and platform access in agentic commerce

Purpose This study develops a theoretical model of how AI shopping agents reallocate shopping activity, reshape seller visibility and alter platform control in digital commerce. Consumer attention is defined as the allocation of shopping episodes across channels, distinct from visual or cognitive attention. Design/methodology/approach A two-seller platform model compares platform-mediated and AI-agent-mediated shopping. The platform chooses agent access and ranking bias; sellers choose price, platform-facing visibility and agent-facing adaptation. Heterogeneous AI-use costs make channel use endogenous, while a fixed implementation cost creates an adaptation threshold. Extensions separate information access from transaction execution and test the robustness of the analytical regimes numerically. Findings Agent access is profitable when transaction and AI-channel gains outweigh lost ranking monetization and governance costs, and information access may remain open while transaction execution is restricted. Sellers activate agent-facing adaptation only when its incremental return covers implementation cost. AI-mediated shopping lowers prices when comparison pressure dominates verified differentiation. Ranking mismatch can turn the two visibility investments from complements into competing activities, and privately optimal openness can fall outside the welfare-improving interval. Originality/value Four features distinguish AI shopping agents from platform recommenders and comparison sites: delegated choice authority, cross-platform scope, multi-stage task execution and an independently governed ranking layer. The framework links these features to layered platform access, endogenous AI-channel use, distinct visibility investments, implementation costs, verifiable trust signals and divergence between private and social incentives in agentic commerce.