Consumer Agency and Decision Quality in AI-Assisted Augmented-Reality Commerce

AI assistance and augmented reality (AR) can make shopping easier without necessarily improving the resulting decision. This conceptual article develops a model of when consumers can inspect, redirect, or reject assistance and when those opportunities may fail. A targeted literature synthesis connects marketing research on autonomy and algorithm overdependence with human–AI reliance studies. The model separates effective decision authority from perceived agency, and distinguishes confidence in advice before acceptance from confidence in the final choice. Three central predictions concern cosmetic controls, checking versus delegation, and persuasive presentation that encourages reliance on inferior advice. Shopping-specific extensions examine whether corrected requirements propagate across a product shortlist, conversation, and AR display, and whether repeated claims from one source are mistaken for independent verification. Decision quality is evaluated only where independent product facts and defensible preference benchmarks permit comparison with an initial unaided choice. A staged preference-review protocol identifies ambiguous cases rather than assigning them arbitrary correctness labels. The evidence table records access depth, and the research agenda specifies results that would count against the central predictions. Practical and governance implications are proposals to evaluate. The model is unvalidated and makes no claim to demonstrate consumer-welfare effects.