Using the FACES stimuli (https://faces.mpdl.mpg.de/imeji/) we compared the extent of automatic processing of age, gender and facial expression. In different blocks of 200 face stimuli participants were asked to either categorize face photos by age (rather young vs. rather old), gender (rather female vs. rather male) or emotional expression (rather emotionally expressive vs. rather neutral). The participants were presented with blocks of the three different tasks in random order. For each task there were two blocks of 200 stimuli: one with happy vs. neutral faces and one with sad vs. neutral faces (counterbalanced across participants). Overall each participant worked on 6 blocks (1200 trials total). After excluding 2 participants with an error rate >10%, we analyzed the data of N=90 participants of this within-subjects laboratory experiment. In order to tap into the extent of automatic processing, we focused on RT effects of repetition vs. change of task-irrelevant features from trial n-1 to the current trial. As a main effects we obtained RT costs if gender changed or if age changed as task-irrelevant feature between the preceding and the current trial. Repetition vs. change of emotional expression as task-irrelevant feature did moderate RT in interaction with the repetition vs. change of the task relevant feature.