Why AI literacy amplifies personalization but not transparency in shaping perceived autonomy

Purpose This study investigates why AI literacy differentially moderates the effects of transparency and personalization on perceived autonomy in AI-based tourism recommendation services. Integrating self-determination theory and the stimulus–organism–response model, it examines the asymmetric moderating role of AI literacy across two pathways linking AI service attributes to behavioral outcomes through perceived autonomy. Design/methodology/approach Data were collected from 322 South Korean users who had used AI travel planning tools within the preceding three months. CB-SEM with AMOS 23.0 was employed for structural path analysis, and bootstrapping with 2,000 resamples tested mediation effects. Interaction terms were mean-centered to address multicollinearity in moderation analysis. Findings Transparency and personalization positively influence perceived autonomy, which more strongly affects continuance use intention (ß = 0.49) than satisfaction (ß = 0.33). AI literacy strengthens the personalization–autonomy link (ß = 0.11, p < 0.05) but not the transparency–autonomy link (ß = −0.02, n.s.), revealing asymmetric moderation. AI literacy also exerts the strongest direct effect on perceived autonomy (ß = 0.36), surpassing both service design attributes. Research limitations/implications The cross-sectional design limits causal inference; future research should employ longitudinal or experimental approaches. The Korean-only sample restricts cultural generalizability. Only autonomy among SDT’s three basic needs was examined; future models should integrate competence and relatedness. Practical implications AI platforms may benefit from designing differentiated autonomy-support strategies based on user AI literacy levels. Transparency features should be universally accessible, while personalization features may be more effective when adapted to varying literacy levels. These findings suggest that investments in AI literacy education – such as onboarding tutorials and algorithmic explainers – may strengthen personalization effectiveness, though causal validation through experimental or longitudinal designs is needed before treating these as confirmed prescriptions. Social implications As AI increasingly mediates consumer decision-making, ensuring that users maintain genuine psychological autonomy becomes a societal concern. This study demonstrates that AI literacy empowers users to convert personalized recommendations into autonomous choices, suggesting that public AI literacy education initiatives can help prevent algorithmic dependence and support informed, self-determined consumer behavior. Originality/value This study provides initial empirical evidence for asymmetric moderation by AI literacy across distinct AI service attributes – transparency and personalization – suggesting that transparency operates at an intuitively accessible informational level while personalization requires evaluative processing facilitated by AI literacy. These findings offer a theoretically grounded basis for further investigation with parallel operationalizations across diverse AI service contexts.