Although the link between exercise and body image is well documented, the considerable inter-individual variability in this relationship has been largely ignored. Therefore, the aim of the present study was to test the contributions of key body image and exercise-related moderators (age, body mass index (BMI), exercise frequency, trait body dissatisfaction, internalisation of appearance standards, and body surveillance tendencies) in predicting variability in the exercise-body satisfaction relationship. Thirty-seven undergraduate women completed a questionnaire containing the above trait-based measures and then carried a Personal Digital Assistant (PDA) for a 7-day period. Participants were instructed to use the PDA to self-report their state body satisfaction immediately following an exercise session and also when the PDA signalled at each of six random intervals throughout the day. Multilevel modelling revealed a bi-directional relationship between exercise and state body satisfaction. Moreover, post-exercise increases in state body satisfaction were strongest for individuals who were younger and engaged in regular exercise, and weakest for individuals with high BMI and/or the tendency to compare their appearance with others. These findings highlight potential limits on the efficacy of exercise-based therapy for body image disturbances.