Self-Disclosure to AI and Stress: A Cross-Lagged Panel Study with Social Anxiety as a Moderator

As AI integrates into everyday communication, users are increasingly engaging in self-disclosure to AI. Understanding the dynamic relationship between self-disclosure to AI and stress is important for clarifying implications for users’ well-being. Drawing on the fever model as well as the “rich-get-richer” and “poor-get-richer” frameworks, this study employed a cross-lagged panel model (CLPM) in a two-wave longitudinal survey of 845 Chinese adults over a two-month interval to examine this relationship and the moderating role of social anxiety. Results showed that earlier stress did not predict subsequent disclosure to AI, whereas earlier self-disclosure to AI was associated with lower subsequent stress. Social anxiety did not moderate the effect of prior stress on later disclosure, but weakened the negative association between self-disclosure to AI and subsequent stress. These findings extend the fever model to AI contexts and inform the design of AI systems that support personalized and equitable stress regulation.