Abstract Hallucinations constitute an intriguing model of how percepts are generated and how perception can fail. Here, we investigate the hypothesis that an altered perceptual weighting of the spectro-temporal modulations that characterize speech contributes to the emergence of auditory verbal hallucinations. Healthy human adults (N = 168) varying in their predisposition to hallucinations had to choose the “more speech-like” of two presented ambiguous sound textures and give a confidence judgment. Using psychophysical reverse correlation, we quantified the contribution of different acoustic features to a listener’s perceptual decisions. Higher hallucination proneness covaried with perceptual down-weighting of speech-typical, low-frequency acoustic energy and prioritizing of high frequencies. Remarkably, higher confidence judgments in single trials depended not only on acoustic evidence but also on an individual’s hallucination proneness and schizotypy score. In line with an account of altered perceptual priors and differential weighting of sensory evidence, these results show that hallucination-prone individuals exhibit qualitative and quantitative changes in their perception of the modulations typical for speech.