When actions contingently yield specific effects, bi-directional action-effect associations are formed. According to ideomotor theories of action control, these bi-directional action-effect associations allow us to select appropriate actions based on anticipated, desired effects. Furthermore, the anticipation of future effects leads to anticipatory saccades towards their expected, future location, reflecting a proactive effect monitoring process. Such anticipatory saccades have consistently been shown in very predictable action-effect contexts. Here, we investigated whether anticipatory saccades can also be observed in a more volatile setting in which participants are relatively uncertain about the correctness of their response and, correspondingly, about the to-be-expected effect. Using a (supposed) perceptual discrimination task, we systematically either frequently presented correct feedback or frequently presented incorrect feedback in different phases of the experiment to induce corresponding response accuracy expectations. Furthermore, on 25% of the trials, participants additionally rated their confidence in having answered correctly/incorrectly before being presented with the effect. Overall, we observed relatively few anticipatory saccades in this volatile setting. Importantly, however, anticipatory saccades scaled with participants’ confidence ratings, indicating that uncertainty is accounted for in proactive effect monitoring on a single-trial level.