AI-redesigned packaging and consumer choice at retail shelves

Purpose This study examines whether AI-redesigned packaging creates measurable retail value when competing with original packages on a physical retail shelf. It addresses the need for ecologically grounded evidence by testing AI-redesigned packages as point-of-purchase retail stimuli rather than as isolated digital designs or intention-only visual outputs. Design/methodology/approach A within-subject real-shelf experiment recorded 72 participants. After predefined biometric quality exclusions, the final analytical sample included 64 participants and 512 product-level observations. Participants viewed eight physical coffee packages simultaneously: four original market packages and four AI-redesigned counterparts. Shelf positions were counterbalanced through Latin Square rotation. Mobile eye-tracking, complementary frontal-response indicators, self-report evaluations and a consequential take-home product-selection task were analysed using source-level contrasts, selected-versus-unselected comparisons, participant-clustered logistic models and participant-fixed conditional logit robustness checks. Findings AI-redesigned packages showed faster visual entry, stronger evaluation profiles and a moderate behavioural advantage. They accounted for 142 of 256 selections (55.5%), compared with 114 selections for original packages (44.5%). The advantage remained statistically reliable in participant-clustered and participant-fixed choice models, but it was not universal; the largest participant segment selected an equal number of original and AI-redesigned packages. Originality/value The study contributes to retail and distribution management research by repositioning AI-redesigned packaging as a testable point-of-purchase shelf intervention. It shows that AI redesign can generate a moderate and heterogeneous choice advantage when it improves shelf entry, processing fluency and evaluation, but it does not support a universal AI-superiority claim.