From confidence to commitment: AI self-efficacy and loyalty in mobile fitness applications through a longitudinal study

Purpose With the rise of artificial intelligence (AI) technologies, mobile fitness applications (MFAs) increasingly offer personalized coaching, adaptive training, and real-time feedback. While users' AI self-efficacy and loyalty are crucial for maximizing these benefits, little is known about how AI self-efficacy influences loyalty in this context. To fill this gap, this study draws on cognitive appraisal theory and social support theory to explore how AI self-efficacy affects perceptions of informational, emotional, and appraisal support and how these support types drive user loyalty to AI-enabled MFAs. Design/methodology/approach This study employs a longitudinal survey-based research design, collecting data from 865 respondents. The data were analyzed via the partial least squares method. Findings AI self-efficacy positively influences users' perceptions of informational, emotional and appraisal support. In turn, these three types of social support enhance user loyalty to AI-enabled MFAs and fully mediate the relationship between AI self-efficacy and loyalty. The decomposition analysis reveals differentiated effects: assistance and anthropomorphic interaction primarily strengthen informational support, while comfort with AI and technological skills have a greater impact on appraisal support. Furthermore, users' exercise experience with AI-enabled MFAs moderates these effects–novice users respond more to informational and emotional support, whereas experienced users rely more on appraisal support. Originality/value This study contributes to the literature by revealing how AI self-efficacy and different forms of social support (informational, emotional, and appraisal) collectively shape user loyalty. The findings offer practical guidance for developers to design applications that strengthen user engagement and promote sustained, long-term usage.