ObjectiveWork automation is increasing worldwide, and the probability of job automation has been associated with workers' adverse health outcomes. This study aimed to examine the association of occupation‐level automation probability with work stress and workers' health.MethodsWe used data from a national survey of 14,948 randomly selected general workers conducted in 2016. Job control and job demand were assessed by the Job Content Questionnaire, and working hours and job insecurity were self‐reported. Health outcomes were measured according to burnout and work‐related injury or disease. We derived automation probabilities for 38 occupational groups and conducted multilevel analyses to examine the associations between occupation‐level automation probability and workers' safety and health after adjusting for psychosocial work conditions.ResultsParticipants working in jobs with a high probability of automation were more likely to have low job control, higher job insecurity, and work‐related injury and disease prevalence; whereas workers in jobs with a low automation probability had higher psychological and physical demands and burnout prevalence. Furthermore, automation probability significantly predicted workers' health after adjustment for demographic characteristics and psychosocial work conditions.ConclusionsWorkers with low automation probability jobs may experience work stress other than that captured by traditional measures of job strain. Organizational approaches to improve employment security and psychosocial conditions are essential for workers' safety and health in the context of increasing job automation.