Integrating large language model-generated culturally adapted verbal suggestions into music therapy for university students: A pilot randomized exploratory study

Previous research has suggested that culturally mismatched interventions may cause resistance and reduce the effectiveness of music-based interventions. This pilot randomized exploratory study examined the implementation and preliminary outcomes of a music-based intervention incorporating large language model (LLM)-generated, culturally adapted verbal suggestions among Chinese undergraduate students. A total of 113 participants were randomly assigned to three multicomponent conditions: (1) Chinese film music accompanied by LLM-generated verbal suggestions involving Eastern imagery, (2) a standardized music therapy recording with conventional verbal guidance, or (3) piano music without verbal guidance. Following a 40-minute cognitively demanding task, participants completed pre- and post-intervention measures of state anxiety, positive and negative affect, and heart rate. Baseline-adjusted ANCOVAs showed no group differences in positive affect, F (2, 109) = 1.03, p = .360, η p 2 = .019, or state anxiety, F (2, 109) = 0.68, p = .511, η p 2 = .012. A significant group effect was found for negative affect, F (2, 109) = 4.60, p = .012, η p 2 = .078; the Eastern imagery group had lower adjusted post-intervention scores than both comparison groups (Holm-adjusted ps = .026). Heart rate also differed across groups, F (2, 109) = 10.56, p < .001, η p 2 = .162, with lower adjusted values in the standardized therapy group than in the other two groups. The findings identify potentially meaningful response patterns but do not establish efficacy or the independent effect of culturally adapted verbal suggestions because the conditions differed in musical material and duration. Larger, controlled trials are required. (ClinicalTrials.gov, NCT07359950).