evidoria

← Back to browse

Good practice Imported

Mind Tutor — A Well-Powered UK RCT Finds No Wellbeing Benefit from an AI Chatbot App

United Kingdom · Oxford · See the United Kingdom profile · See the Oxford profile

Evidence: Randomised controlled trial Top 89% 27/100 · Ask Evidence Copilot about this practice

Two well-powered UK RCTs (N=177, N=240) tested an AI chatbot app meant to boost first-year students' wellbeing. Neither found a significant effect, and each student used the app only twice on average.

177 students
Trial 1 sample size (6-week follow-up)
240 students
Trial 2 sample size (8-week follow-up)
2 interactions
Average app interactions per student
80 students
Intervention-arm students with usable engagement data (acceptability sub-study)
125 students
Control-arm students with usable engagement data (acceptability sub-study)
Mind Tutor — A Well-Powered UK RCT Finds No Wellbeing Benefit from an AI Chatbot App

Details

Maturity
Pilot
Promoter
Oxford Brookes University, London Metropolitan University, University of Birmingham, Syndeo Ltd
Period
2021-2025 (RCT published 2023; development/acceptability paper published 2025)
Keywords
higher education, student wellbeing, mental health, AI chatbot, randomised controlled trial

Context

Mind Tutor is an AI-chatbot-enhanced wellbeing app built for first-year undergraduates, developed by a team from Oxford Brookes University, London Metropolitan University and the University of Birmingham together with the app developer Syndeo Ltd. The chatbot guides students toward wellbeing content addressing anxiety, low mood, the transition to university, and relationships.

Objectives

The research team set out to test whether the AI-chatbot-enhanced app could improve first-year student wellbeing, via two adequately powered, pre-registered randomised controlled trials at two UK universities.

Activities

Trial 1 enrolled 177 students (92 intervention, 85 control, 6-week follow-up) and Trial 2 enrolled 240 students (125 per arm, 8-week follow-up). Sample sizes were large enough to detect even small effect sizes at a 5% error rate, so a Type II error is unlikely to explain the results. The trial was registered in advance on the Social Science Registry.

Results

Neither trial found a significant effect of Mind Tutor on student wellbeing. Engagement was also low: across both trials, each participating student interacted with the app an average of only two times, and only 80 of the original intervention-arm students in the acceptability sub-study provided usable engagement data, versus 125 in the control group.

Conclusions

The research team published both the null efficacy result and the disappointing engagement data openly in peer-reviewed journals — a genuinely evidence-driven, cautionary case about how easily a well-designed AI wellbeing tool can fail to move outcomes once tested rigorously.

Implementation

Indicative cost
Medium (€50k–€500k) — Multi-institution academic RCT plus a commercial app-development partner (Syndeo Ltd); no public budget figure found in the sources reviewed.
Time to results
Long (> 3 years) — Development and trials spanning 2021-2025: efficacy RCT published 2023, development/acceptability paper published 2025.
Staffing & skills
Research team: Oxford Brookes University, London Metropolitan University, University of Birmingham, App developer: Syndeo Ltd

Conditions for success

  • Adequately powered, pre-registered trial design (Social Science Registry) able to detect small effect sizes
  • Open publication of both efficacy and engagement/acceptability results in peer-reviewed journals

Common failure modes

  • Very low real-world engagement — an average of only two app interactions per student — likely undermined any chance of a wellbeing effect
  • Neither adequately powered RCT found a significant wellbeing benefit
  • High attrition in the acceptability sub-study engagement data (only 80 of the original intervention-arm students provided usable data)

Where it fits

Governance type
university research consortium with a private app developer
Scale
two UK university sites
Income level
high-income

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful