"It just happened to be the perfect thing": Real-life experiences of generative AI chatbots for mental health

Abstract

The global mental health crisis underscores a critical need for accessible and effective interventions. Generative artificial intelligence (AI) chatbots, such as ChatGPT, are emerging as a novel solution, but research into their real-life usage is limited. We interviewed nineteen individuals about their experiences of using generative AI chatbots to work on their mental health. Most participants reported high levels of engagement and positive impacts, including improved mood, reduced anxiety, healing from trauma and loss, and improved relationships. Our analysis resulted in four overarching themes: 1) the value of an ‘emotional sanctuary’, i.e., a safe, validating space that is always available, 2) the ‘insightful guidance’ provided, particularly on the topic of relationships, 3) the ‘joy of connection’ experienced, and 4) comparisons between the ‘AI therapist’ and human therapy. Some of these themes echo previous research on rule-based chatbots, while others appear to be novel to generative AI. Participants highlighted the need for a better approach to safety guardrails, more human-like memory and the ability to lead the therapeutic process. Our findings suggest that generative AI chatbots may offer meaningful mental health support, but further research is needed to explore their safety and effectiveness.