When Looks Matter: How Virtual Agent Attractiveness and Appearance Satisfaction Shape Self-Disclosure in the Metaverse

As artificial intelligence (AI)–based virtual agents become increasingly embedded in immersive environments, understanding how their design influences user behavior is essential. Drawing on the Computers Are Social Actors (CASA) framework and social comparison theory, this study investigates whether virtual agent attractiveness affects self-disclosure in a metaverse-based health consultation and how this effect depends on users’ appearance satisfaction. In a between-subjects experiment with 163 female participants, we manipulated virtual agent attractiveness (high vs. low) and measured appearance satisfaction. Moderation analysis revealed that attractiveness reduced disclosure intentions among individuals with low appearance satisfaction but had no significant effect among those with average or high satisfaction. These results suggest that virtual agent attractiveness is not a universally positive cue but interacts with self-perceptions of appearance to shape disclosure behavior. The findings extend CASA and Proteus Effect perspectives and offer practical implications for the design of socially intelligent and ethically responsive virtual agents in sensitive contexts.