Exploring trust, emotional disclosure, and help-seeking among university students engaging with generative AI chatbots for mental health support

Abstract

In recent years, university students have increasingly used generative AI chatbots for emotional expression and informal mental health care. This qualitative study investigates Pakistani university students' experiences, perceptions, and trust in emotional disclosure and help-seeking with generative AI chatbots. It also explored the issues students raised with the chatbots, why students felt they preferred to seek AI support over human support, and the perceived advantages, dangers, and moral considerations of such practice. The participants who were selected for the study were 22 university students who had used generative AI for emotional or mental health purposes through purposive sampling. Semi-structured interviews were conducted to gather data, which was then analysed using Braun and Clarke's reflexive thematic analysis. Five themes were uncovered: trust was developed gradually and remained conditional; students disclosed concerns they could not easily share with people; chatbots were experienced as available and nonjudgmental; AI acted as both a bridge and a barrier to human help-seeking; and privacy, accuracy, cultural relevance, and safety concerns limited trust. The results indicated that students valued the immediate availability, emotional privacy, and supportive language of AI chatbots. They didn't think AI is equally helpful as mental health professionals. Generative AI's role as a tool that complemented emotional reflection and preparation for human assistance was most helpful. Universities must provide culturally competent advice regarding safe use of AI, privacy, crisis support, and referrals to professional sources.