College students and large language models for mental health: A qualitative exploration of motivations and experiences

Park, J. J., Brown, S. A., Chu, S. L., Zhou, A. Q., & Abreu, R. L. (2026)

International Journal of Human-Computer Studies

https://doi.org/10.1016/j.ijhcs.2026.103806

College students face significant mental health challenges, yet traditional support systems often remain inaccessible due to stigma, scheduling constraints, and limited institutional resources. Large Language Models (LLMs), such as ChatGPT, offer a promising avenue for providing immediate, anonymous, and personalized mental health support. This study employed a descriptive phenomenological approach to explore how and why college students engage with LLMs for mental health purposes, examining their motivations, lived experiences, and perceived benefits and limitations. 20 students were interviewed about their use of LLM-based tools such as ChatGPT, Pi, and Character AI. Findings revealed broad themes encompassing motivations, experiences, strengths, and limitations. Students were motivated by curiosity, stigma avoidance, and the need for accessible, self-directed support, while their experiences reflected emotional, relational, and psychoeducational engagement with LLMs. They described LLMs as accessible, nonjudgmental, and empowering tools that provided validation, comfort, and practical coping strategies within the constraints of student life. Yet, these positive experiences coexisted with contradictions. For example, students simultaneously perceived LLMs as empathetic and mechanical, safe yet uncertain, and personalized yet impersonal. Despite these tensions, LLMs ultimately functioned as “good enough companions” for college students, offering meaningful, immediate support while revealing the limits of artificial empathy. The study highlights design implications for developing specialized, transparent, and ethically governed LLM systems to complement traditional mental health services in university settings.