We propose that individuals incorrectly interpret their success in predicting an event as an indication that they understand the event's underlying mechanism or process. We term this illusion prediction-comprehension bias (PCB) and conceptually locate it within the framework of reverse inference logic. In a set of five experiments (N = 851), we demonstrate this bias and identify three psychological factors that produce it: The feedback participants receive about their prediction accuracy (Experiment 1), the consistency with which a cause and its effect are presented together (Experiment 2), and the fluency of an observation brought about by the number of times it is repeated (Experiment 3). In Experiment 4, we directly replicate the results of Experiment 3 and explore the bias’s underlying mechanism: We find first evidence for that respondents use the perceived validity of inferring prediction from comprehension (a sound inference) as an indicator for when to draw the reverse inference (inferring comprehension from prediction) that underlies the PCB. In Experiment 5, we corroborate the PCB-mechanism by demonstrating that experimentally manipulating the validity of inferring prediction from comparison affects how strongly participants exhibit the PCB. Failed reverse inference logic seems to substantially account for the PCB.