Millions of teenagers have turned AI into their go-to psychologist. It is an unprecedented challenge for medicine

In society there is a fairly well-established debate about how social networks affect the mental health of young people and there is even debate about the possible consequences they have, going so far as to propose very clear limits for accessing them. However, while the focus was on the recommendation algorithms of TikTok or Instagram, a new trend has been quietly growing on the screens of millions of teenagers: the use of generative AI as a therapist.

New therapies. Here, research led by the RAND Corporation has brought to light the magnitude of this phenomenon by analyzing a sample of 1,058 young people between 12 and 21 years old. And the figures paint a quite revealing picture by pointing out that 13.1% of adolescents and young adults use generative artificial intelligence to obtain advice about their mental health.

But the most worrying thing is that this percentage shoots up to 22.2% if we look exclusively at the oldest age group, that is, those between 18 and 21 years old. And although it can be defended as something specific, the reality is that 65.5 of these users turn to AI on a monthly or even greater frequency.

Works? The most striking thing we have learned from this study is not only that young people consider AI as a psychologist, but that those who come leave quite happy, since 92.7% of users stated that the advice provided by AI was useful to them.

And among the reasons they give for their satisfaction, what stands out above all is the possibility of resorting to their 'services' at any time, the absence of economic barriers and, above all, the feeling of privacy and lack of human judgment. All of this together is turning large AI models into the first line of emotional support for Generation Z.

The other side of the coin. Just because a tool is perceived as useful by the user does not mean that it is clinically safe, because the intersection between generative technology and psychiatry is a minefield, and major medical institutions are already raising their hands.

In summer 2025, the American Psychological Association issued an official warning about the risks of relying on AI for the diagnosis or treatment of mental disorders. Among the reasons they give, it stands out that language models are designed to predict the most likely next word and sound empathetic and convincing, but they lack real understanding, clinical context and the ability to manage severe crises.

The security. Added to this warning is the devastating context provided by researchers at Stanford University, who also in 2025 evaluated the responses of several chatbots to mental health queries. Their conclusion was worrying as they saw that in 1 in 5 cases, the artificial intelligence provided advice that was unsafe or inappropriate for the user's situation.

A real challenge. Right now we are at an inflection point where AI is filling a huge gap in a mental health system that, globally, is collapsed and inaccessible for a large part of the young population. And furthermore, prohibiting or blocking access to these tools does not seem like a realistic solution in the face of millions of users who have already integrated them into their emotional well-being routine.

That is why the real challenge for technology companies and health agencies is twofold: on the one hand, improving the security barriers of the models so that they refer users to human emergency services when necessary.

Images | Daria Nepriakhina 🇺🇦

In Xataka | There is a weapon of mass destruction against our ability to remember things: stress