Music plays a central role in psilocybin therapy (PT), where playlists are designed to be supportive and stimulating over a 4–6-hour psychedelic experience. The primary design principle underlying PT music programming is that music should be affectively aligned with the onset–peak–return model of pharmacodynamic action. However, in the absence of clear acoustic criteria, this framework has been implemented largely through individual or team-based curatorial judgment, making it unclear whether contemporary research PT playlists consistently realize the proposed onset–peak–return progression. Here, we test this question by analyzing 45.1 hours of audio from eight PT music playlists (369 tracks) for systematic acoustic and affective organization across onset, peak, and return phases. To this end, we applied supervised and unsupervised machine learning approaches to a comprehensive set of audio features extracted from PT music playlist tracks, evaluating phase differentiation across the full dataset as well as within individual playlists. The results demonstrate that although weak phase-based differentiation is apparent for low-level audio features, especially within some playlists, the overall set lacks consistent phase-based structure. This implies that current curatorial practices are not aligned with their theoretical basis, raising questions about how such practices can evolve to better support emerging psychedelic treatments.
