Purpose Amid the rapid growth of the digital economy and rising social isolation, companion Artificial Intelligence (companion AI) has evolved from traditional conversational agents into sophisticated, emotion-oriented interaction systems that address users' socioemotional needs. This study examines how companion AI influences users' willingness to sustain interaction, identifies the mechanisms underlying this effect, and explores the AI attributes and consumer characteristics that moderate these relationships. Design/methodology/approach Grounded in anthropomorphism theory and construal level theory, four controlled experiments examined how AI role type (with a personal role vs. with no role) influences users' interaction intentions. The model tests psychological distance as a mediator, with AI empathy and users' self-concept clarity as moderators. Findings Results show that companion AIs with personal roles increase continuous interaction intention. Psychological distance mediates this effect, while high empathy reduces distance and strengthens engagement. Users with lower self-concept clarity show greater reliance on personal-role AIs for emotional connection. Originality/value This research fills a gap in companion AI studies by extending role theory to human–AI interaction and revealing how AI role cues shape engagement. It advances theory building and provides practical guidance for personalization and ethical AI design.
Fan et al. (Fri,) studied this question.