Young people use ChatGPT, Gemini and Claude as therapists: the problem isn't immediacy, but mistaking it for a bond

By TV4 Noticias Guanajuato · November 7, 2026.

ChatGPT, Gemini and Claude are no longer used only to answer questions or complete tasks, but also to seek advice and emotional support. According to the report, a growing number of young people are turning to them as an alternative to psychological therapy without asking how safe it is to trust them with their mental health.

The source is Karina Márquez, coordinator of the Psychology degree program at Universidad Iberoamericana León. She explains that the main appeal is the immediate response: there is no need to wait a week or two for an appointment, or to spend money, time and travel.

Márquez points to two risks. The first is that tailoring responses to what the user says creates a sense of closeness, but not a therapeutic bond or personalized care. The second is self-diagnosis: a diagnosis requires interviews, observation and reliable psychological tests, and a misinterpretation can have consequences in the short, medium or long term.

She also acknowledges benefits when the tools are used responsibly: looking up information, finding relaxation exercises, accessing educational materials or tracking mood with apps. Her advice is not to rely entirely on the responses, not to share personal information, and to remember that AI is no substitute for professional care.

Our take: the article is the view of a local expert, not a study, and it offers no figures on how many young people do this or on specific harms. It should not be given more weight than it deserves. Even so, it points to something real: people do not turn to these chatbots out of naivety, but because mental health services arrive late and cost money. Blaming the user does not address that cause.

Our position is that the useful question is not "does it replace therapy or not?" but what role it can play without causing harm. For context in the sector, this ties into a pattern we see in healthcare: public acceptance of AI depends on human oversight, and the risk lies as much in consent and data custody as in the model's accuracy. The same applies here: the warning not to share personal information is not a minor detail; it is half the problem.

In the short term, there is an uncomfortable transition: millions of people will use these tools as their first emotional confidant before there are clear rules, safety standards or referral pathways to a professional. The winners will be developers who design explicit limits, such as detecting crises and referring users onward; the losers will be those who sell the illusion of full-fledged therapy.

In the long term, we see an opportunity. An AI that offers a first point of listening, psychoeducation and mood tracking, integrated with human therapists, could expand access to mental health care for people who cannot reach it today. This is consistent with our abundance thesis: that quality care should stop being a privilege of time and money. But that future requires clinical judgment to remain human and the system not to use the chatbot as an excuse to cut therapists.