Filling the Gap? Young People’s Use of AI for Mental Health Support

Artificial intelligence (AI) is changing how children seek help with their mental health and wellbeing. The policy question is no longer whether children will use AI for support, but how we ensure that the technology they are already using is safe, effective, and does not become a substitute for the human relationships and the services that children want and need.

To gain a greater understanding of children’s experiences and expectations, Barnardo’s and the National Counselling and Psychotherapy Society (NCPS) commissioned Censuswide to survey 2,000 children aged 11 to 17 across the UK. The findings reveal a rapidly developing reality: 43% of children said they had used a mental health app or chatbot for emotional support or advice in the previous year. We found that children are using both general-purpose chatbots, and purpose-built tools to support their mental health, often alongside each other. Children are therefore already using AI in ways for which much of the technology, and the system around it, was not designed.

But children are also clear about what good mental health support looks like and the limitations of both general-purpose and even purpose-built AI models in delivering this. When asked who they would want to talk to if they were going through something difficult, 59% chose family or caregivers and 18% friends, while just 4% chose a chatbot. Almost 90% said trust and feeling understood and cared about were important to good mental health support, while 57% did not believe AI could provide empathetic support.

Our research exposes an important gap between the support children want and the support they can access. Of the 43% of children who had tried to access professional mental health support in the previous year, only 27% received the help they needed. That gap is particularly stark for children from lower-income households. 24% of those living in households earning under £15,000 described themselves as on a waiting list, compared with around 11% in the highest- income group. Yet children from lower-income households were also less willing to consider a purpose-built AI tool for their mental health and more likely to be unhappy about its increased use by schools, workplaces or the NHS. The choice facing policymakers is therefore not between embracing AI and banning it. It is about determining the role we want AI to play in children’s mental health and putting the protections in place to make that role safe.

Barnardo’s and NCPS believe that general purpose AI has the potential to provide useful information, practical support and routes into human help, while purpose-built AI could sit alongside other interventions, supporting referral, triage and self-management under the supervision of clinicians. However, all use of AI tools for mental health should complement, not replace, the relational support children tell us they value.