This multistudy report used an experimental approach to alter automatic evaluations of exercise (AEE). First, we investigated the plasticity of AEE (study 1). A computerized evaluative conditioning task was developed that altered the AEE of participants in two experimental groups (acquisition of positive/negative associations involving exercising) and a control group (η2 part. = .11). Second, we examined connections between changes in AEE and subsequent exercise behavior (chosen intensity on a bike ergometer; study 2) in individuals that were placed in groups according to their baseline AEE. Group differences in exercise behavior were detected (η2 part. = .29). The effect was driven by the performance of the group with preexisting negative AEE that acquired more positive associations. This illustrates the effect of altered AEE on subsequent exercise behavior and the potential of AEE as a target for exercise intervention.