
Researchers from Black Dog Institute, UNSW and Deakin University have completed a world-first clinical trial using artificial intelligence (AI) to determine which digital mental health tools work best for university students experiencing psychological distress.
Published in JAMA Network Open, the study tested four self-guided, smartphone programs—physical activity, mindfulness and sleep hygiene and an active control—in more than 1200 university students who reported elevated distress.
Using a world-first AI-enhanced adaptive trial design, the research team tested multiple app-based programs at once and rapidly identified which delivered the greatest benefits for different groups of students—cutting months off traditional clinical trial timelines.
“This is the first time AI has been used to match digital mental health tools to students’ individual needs,” Black Dog Institute and UNSW Professor Jill Newby said.
“We were able to learn, in real time, which brief programs worked best for different levels of distress—so we can deliver faster, more effective and personalised support.”
The trial was delivered entirely through the Vibe Up smartphone app. Participants—each currently enrolled in tertiary education—received remote access to brief, psychoeducational interventions designed to fit into their daily routines. The study ran a series of AI-powered ‘mini-trials’, each lasting up to one month, to identify which treatments worked best for different levels of distress.
The study found that mindfulness and physical activity worked best for students with severe distress, while sleep hygiene and physical activity were most effective for those with mild distress. For moderate distress, no single approach stood out, highlighting the need for further tailored options.
Traditional clinical trials often take years and require large control groups. In contrast, this adaptive design used real-time data to direct participants to the most promising interventions as results emerged. Researchers estimate a conventional trial would have required 25 per cent more participants to detect the same effects—demonstrating the power of AI to make mental health research faster and more efficient.


