“It’s Often Feeling Nothing...”: Evaluating LLMs for Mental Health Literacy Interventions with College Students

Mental health is a critical aspect of overall well-being, yet stigma and misinformation remain prevalent, particularly among college students who face unique stressors. The rising incidence of mental health conditions, exacerbated by the COVID-19 pandemic, underscores the need for accessible support systems. Digital mental health interventions, such as narrative-based games, offer a scalable, affordable, and engaging approach to promote mental health education. This study combines visual novels with Large Language Models (LLMs) to create a digital intervention aimed at improving mental health literacy by addressing literacy on depression and misconceptions about depression. We developed a visual novel using LLM-generated narratives and conducted a mixed-methods study with 28 college students, assessing the game’s impact at pre-test, post-test, and one-week follow-up, along with narrative transportation. Results showed no significant change in depression literacy, likely due to high baseline scores, but misconceptions of depression significantly decreased and were maintained at follow-up. Participants reported moderate to high narrative transportation. Qualitative findings emphasised emotional engagement, stigma reduction, the perceived value of digital tools, and participants’ scepticism regarding AI-generated narrative authenticity.