Teenagers are increasingly turning to AI chatbots as confidants, coaches and, sometimes, as a substitute for therapy. Most of these systems were not designed, tested or regulated to provide mental health care.
The evidence base is still small and sometimes mixed, but it is not reassuring. In a study of 1,131 U.S. adults using Character.AI, researchers at Stanford and Carnegie Mellon found that people with smaller social networks were more likely to report using chatbots primarily for companionship. Companionship-oriented use was associated with lower wellbeing, particularly when interactions were more intensive, more personally disclosive, and when users had less human social support.
Other research has found that publicly available chatbots can miss serious distress, respond inconsistently to signs of suicide risk, and give advice that may be unsafe for vulnerable users. This matters especially for adolescents, whose distress can be indirect, ambiguous, or disclosed slowly over time - exactly the kinds of signals that current systems can struggle to interpret.
We do not start from the position that this technology is bad for young people. There are nowhere near enough trained mental health professionals to meet global need, and well-built tools could extend access to evidence-based support, triage and early intervention for people who currently have none.
The question is not whether young people will use AI for emotional support. Many already do. The question is whether anyone is studying that use closely enough, and building these tools carefully enough, for them to be safe.
The children using these systems today cannot wait for the evidence base to mature. Nor should they have to rely solely on the companies building the tools to get it right. That is why part of this portfolio goes into classrooms helping young people understand not only how AI systems work, but who builds them, how those companies make money, and who young people are becoming in relation to technologies designed to keep them engaged. That kind of agency is not a substitute for safer design or better regulation. It is the part that travels with a child into every tool they will use for the rest of their life.
Our portfolio funds both sides of the question: understanding what is happening now, and helping to change what happens next.
In the coming days, we will introduce the research, organisations and programmes we are supporting and explain why we believe this is one of the most urgent open questions in children’s wellbeing.