This study investigates the causal impact of subjective AI-induced job replacement anxiety on work motivation, turnover intentions, and mental well-being among Japanese workers. Integrating task-based theory, efficiency wage theory, and health capital models, we formalize the fear of future replacement as a microeconomic friction. To address endogeneity, we employ a simultaneous ordered probit framework using the ILO’s Global Index of Occupational Exposure and Japan’s Automation Risk Index as instrumental variables. Results reveal that AI anxiety significantly erodes work engagement and meaningfulness. Critically, AI anxiety heightens turnover intentions for the full sample, strictly aligning with theoretical predictions. However, we identify a non-monotonic generational pattern: middle-aged workers exhibit a defensive “lock-in” effect, whereas younger and older cohorts show increased mobility. Furthermore, mental health deterioration is concentrated among younger and self-employed individuals. These anticipatory psychological frictions distort labor incentives long before actual displacement, highlighting the need for human-centered AI policies.