Convenience or risk? Understanding users' continuance intention toward AI health assistants.

In recent years, AI health assistants have rapidly proliferated in the field of personal health management, yet systematic explanations of the mechanisms underlying users' continuance intention remain lacking. Building upon the Expectation-Confirmation Model (ECM), this study integrates variables including task-technology fit (TTF), perceived risk, AI literacy, perceived value, and AI hallucination to construct an extended model that explores the relationships among confirmation, perceived usefulness, satisfaction, and continuance intention. This study employs a two-stage research design: Study 1 collects 700 valid samples through online questionnaires and utilizes structural equation modeling (SEM) to test hypothesized relationships and the moderating effects of gender differences, while Study 2, based on 300 samples, conducts exploratory factor analysis (EFA) combined with open-ended questionnaire data to further identify key driving and inhibiting factors affecting continuance intention. The results indicate that confirmation significantly enhances perceived usefulness and satisfaction, with satisfaction further promoting continuance intention. Additionally, TTF and AI literacy exert significant positive effects on both satisfaction and continuance intention, whereas perceived risk demonstrates significant negative effects on confirmation and satisfaction, and AI hallucination negatively influences satisfaction. Gender exhibits significant moderating effects on the pathways through which TTF, satisfaction, and AI literacy affect continuance intention. Study 2 further reveals that factors such as emotional engagement, cognitive support, tool effectiveness, and perceived ease of use constitute important facilitators of continued use, whereas technical burden, emotional detachment, and resource constraints represent major barriers. This study not only validates and extends the applicability of ECM in the AI health assistant context but also provides theoretical foundations and practical insights for product optimization and differentiated promotion of AI health assistants.