In everyday life, we use perception to guide our behaviour. While much effort has been devoted to neurophysiologically study perception and behaviour in isolation from each other, studies that conjoin perception and behaviour are rarer. Here, we devised a novel paradigm to dynamically study the action-perception loop, framed in terms of predictive processing as a guiding framework for brain function. We tracked the electrophysiological markers of predictive processing by using hierarchical frequency-tagging in an active foraging and recognition task. Participants had to forage a two-dimensional landscape to find three target images. They freely selected their foraging paths and when to finish and move to the next landscape. Temporally resolved analyses of hierarchical frequency-tagging signals revealed that putative prediction error signals triggered a cascade of neural signalling events leading to recognition. In addition, our results show that the accumulation of uncertainty is correlated with the decision to abort foraging and start a new search. For the first time, we tracked temporally-resolved frequency-tagged signals in an action-perception paradigm; this is consistent with contemporary iterations of predictive processing that increasingly focus on action (active inference). Our paradigm and findings open new ways to study such signals during the action-perception cycle beyond passive settings.