Conversational AI or a human professional for mental health advice? Exploring prevalence and public preferences in Australian adults

Abstract

Objective: General-purpose conversational AI (e.g., ChatGPT) is increasingly accessed for mental health support despite lacking clinical validation. This study examined the prevalence and user preferences for conversational AI compared to mental health professionals among Australian adults. Method: A two-phase survey recruited 500 Australian adults (215 male, 278 female, 7 gender diverse) via an online crowdsourcing platform. Phase 1 assessed the prevalence of AI use for mental health. Phase 2 invited participants with dual experience of AI and professional care for mental health (n = 82) to rate comparative preferences across trust, competence, accessibility, and satisfaction, and indicate their preferred modality for future mental health support. Results: Approximately 24% of participants had discussed mental health with AI. In direct comparison, AI was strongly preferred for accessibility, while professionals were rated significantly higher for trustworthiness, satisfaction, and most competence items. The majority (52.4%) preferred using both modalities for mental health support, though 19.5% preferred AI exclusively. Conclusion: Users demonstrate clear preferences: professionals for clinical quality, AI for accessibility. A majority prefers accessing both modalities for mental health support, yet current systems provide no framework for integrating or supervising such hybrid use. These findings suggest mental health services should consider how to guide rather than ignore clients’ AI engagement.