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

In a novel intersection of artificial intelligence and music therapy, a recent pilot randomized study has examined the effects of culturally adapted verbal suggestions generated by large language models (LLMs) within music-based interventions for Chinese university students. Conducted in early 2026, the investigation aimed to address challenges posed by cultural mismatches in therapeutic settings, which have been shown to hinder treatment acceptance and diminish outcomes.

The study enrolled 113 Chinese undergraduates aged 18 to 25, who were randomly divided into three groups. The first group experienced a music therapy session combining Chinese film soundtrack excerpts with verbal suggestions crafted by an advanced LLM to incorporate Eastern cultural imagery and symbolism. The second group received a standardized music therapy program featuring conventional verbal guidance without cultural tailoring. The third group listened to piano music without any accompanying verbal instructions.

Participants underwent a 40-minute cognitively demanding task before interventions and were assessed on psychological metrics, including state anxiety, positive and negative affect, alongside physiological responses measured by heart rate. The intervention featuring culturally adapted verbal suggestions utilized music from Tan Dun’s soundtrack for the 2002 film Hero, known for its strong representation of traditional Chinese aesthetics. Verbal prompts generated through ChatGPT 5.2 were carefully reviewed and refined to ensure alignment with Chinese cultural values and linguistic nuances.

Statistical analysis revealed that while positive affect and state anxiety showed no significant differences across the groups, the culturally adapted verbal suggestion group demonstrated notably lower levels of negative affect post-intervention compared to the other conditions. Interestingly, the standardized therapy group exhibited lower heart rates than those in both the culturally adapted and music-only groups, suggesting distinct physiological responses depending on the intervention type.

These findings highlight the potential benefits of embedding culturally resonant language within music therapy to alleviate negative emotional states in university students, a population increasingly vulnerable to academic and employment-related stressors in the post-pandemic era. However, the study’s authors caution that the results do not establish definitive efficacy or isolate the effects of verbal suggestions alone, due to variations in musical selections and intervention durations among groups.

The research emerges amid growing recognition of mental health challenges facing undergraduates globally, particularly in China, where academic pressures and economic uncertainties have intensified anxiety and burnout. Traditional stress management approaches such as cognitive behavioral therapy and pharmacological treatments face limitations including accessibility and side effects, driving interest in alternative, non-invasive methods like music therapy.

Music therapy’s appeal lies in its safety, affordability, and ease of implementation, particularly valuable within educational settings. Verbal suggestions, a pivotal component of such therapy, enhance emotional engagement and have been shown to regulate both subjective feelings and physiological indicators such as heart rate. Yet, the direct translation of Western therapeutic language often clashes with Chinese cultural communication styles that favor metaphorical and indirect expression, potentially reducing therapeutic rapport.

By leveraging the advanced natural language capabilities of LLMs, this study pioneers a scalable approach to crafting culturally congruent verbal guidance in music therapy. The experimental design included meticulous alignment of verbal prompts with the emotional contours of the selected film music, structured into phases promoting emotional engagement, tension release, integration, and relaxation. This integration was intended to preserve the immersive quality of the music experience while enhancing therapeutic impact through culturally meaningful language.

Despite promising initial patterns, the study underscores the necessity for larger, rigorously controlled trials to robustly assess the independent contribution of culturally adapted verbal suggestions and confirm their clinical significance. Data and intervention materials have been made publicly available, inviting further exploration and validation by the research community.

This investigation marks a significant step toward personalized mental health interventions that respect and incorporate cultural identity, leveraging artificial intelligence to enrich therapeutic modalities. As universities grapple with rising student stress and limited resources, such innovations may offer accessible, culturally sensitive tools to support psychological resilience and well-being.