Processing stimuli in the visual periphery is impaired by adjacent stimuli, a phenomenon known as visual crowding. The present study investigates how task type (localization, feature detection, categorization, and target presence detection) and stimulus characteristics (shape and category) affect crowding. Both offline and online experiments were conducted. In two experiments, we used a 2-alternative forced choice (2AFC) paradigm, presenting crowded targets (K, U, 3, 4) flanked by one or three distractors (E, S, 7, 8). Stimuli were presented at 8° eccentricity for 300 ms, with a horizontal spacing of 0.5°. Participants completed four blocks, each corresponding to one task. The block order was counterbalanced across participants. Classical results from crowding studies were observed: participants performed better in the 1-flanker condition compared to the 3-flanker condition. We also observed strong effects of tasks on performance: localization showed the lowest accuracy, while detection yielded the highest. Feature detection and categorization tasks showed similar performance levels. The impact of stimulus features and categories also varied by task. For example, curved targets surrounded by curved flankers led to lower performance in the detection task but higher performance in categorization. Digit targets surrounded by letter flankers led to worse performance in categorization but did not affect detection or localization. Reaction time was fastest in the detection task. Overall, the results show that both the tasks participants are asked to perform and the stimuli they are presented with influence performance in crowded-stimulus perception tasks. Findings are discussed in terms of the hierarchical organisation of visual processing.