Approach-avoidance training (AAT) is a procedure for changing people's likes and dislikes that involves executing repeated approach (e.g., pulling a joystick towards yourself) and avoidance actions (e.g., pushing a joystick away from yourself) in response to target stimuli. Typically, this leads to approached stimuli being evaluated more positively than avoided stimuli. However, the evidence that AAT can change evaluations of feared stimuli is mixed. In this preregistered study, we wanted to investigate the effectiveness of a novel version, compared to a more typical version, of AAT for changing the evaluation of fear conditioned stimuli.After a differential fear conditioning phase, participants (N = 80) were randomly allocated to two conditions: In the novel AAT, participants repeatedly approached one positive picture (i.e., puppies) and avoided one negative picture (i.e., a dead cat) in addition to approaching and avoiding the conditioned stimuli. Participants' evaluations of the stimuli were assessed with explicit ratings and an affective priming task.We found evidence for the effectiveness of approach-avoidance training to change evaluations of fear conditioned stimuli. However, we found no evidence for the superiority of our novel version of the AAT procedure.The sample size of our study was quite small, limiting the statistical power to detect small effects.Both a typical and an adjusted version of the AAT procedure proved successful to change conditioned negative evaluations. We compare our findings to previous studies showing limited effectiveness of the AAT procedure with feared stimuli.