When Trust and Privacy Pull in Opposite Directions: Consumer Decision Making with Conversational AI

Abstract

Drawing on the Stimulus–Organism–Response (S-O-R) framework, this study examines how consumers’ perceptions of information quality, personalization, explainability, and responsiveness are associated with purchase intention through their relationships with perceived usefulness, algorithmic trust, and privacy concern, while considering AI literacy as a boundary condition. Data from 588 Chinese online shoppers who completed a standardized shopping task through direct interaction with a conversational AI shopping assistant were analyzed using structural equation modeling. The results show that all four consumer-perceived conversational-AI cues are positively associated with perceived usefulness and algorithmic trust, whereas perceived personalization is also positively associated with privacy concern. Perceived usefulness and algorithmic trust are positively associated with purchase intention, whereas privacy concern shows a negative conditional association with purchase intention in the structural model. Personalization shows competing indirect associations with purchase intention: positive indirect associations through usefulness and trust coexist with a negative indirect association through privacy concern. The positive trust–intention association and the negative conditional privacy concern–intention association are stronger at higher levels of AI literacy, whereas the usefulness–intention interaction is statistically compatible with practical equivalence relative to the specified study-specific bound. These findings indicate that consumer responses to conversational AI reflect competing benefit–risk evaluations rather than a uniformly positive acceptance process, and that AI literacy is differentially associated with the strength of the trust–intention and privacy concern–intention relationships.