Nearly 3 in 10 Peruvian teenagers confide their troubles to AI: the problem isn't the chatbot, it's the void it fills

By ATV.pe · November 4, 2026.

Nearly 3 in 10 Peruvian teenagers turn to artificial intelligence to vent or seek advice when they feel sad or anxious, according to ATV.pe, which reports on an interview with clinical psychologist Manuel Saravia on ATV Noticias Edición Mediodía. The outlet's text does not specify the source or methodology behind that figure, so it is best read as an order of magnitude rather than a definitive statistic.

Saravia argues that AI can offer emotional containment, meaning the feeling of being heard, but it does not provide therapy or replace parents. According to the specialist, these tools tend to make users feel good, and their sugarcoated answers sometimes keep people from seeing reality. In serious situations, such as suicidal tendencies or eating disorders, he recommends first turning to a trusted adult and seeking a professional right away.

The psychologist also warns of a broader dependence that extends to young people and adults alike: people who consult AI even for simple decisions. In schools, he describes a pattern of identical answers when several students use the same prompt (the instruction written to the model) for an assignment; he mentions classrooms of 40 students where everyone can end up with the same answer. His conclusion is that the point is not to demonize AI but to combine critical thinking with skilled use of the tool, and he believes its use by minors should be regulated.

Our take. The most interesting takeaway from the interview is not technological. That a teenager would prefer a chatbot to their parents or friends speaks to availability (it answers at any hour), the absence of judgment and zero friction. These are real strengths of the tool and, at the same time, shortcomings of the people around them. Saravia puts it well when he asks adults to consider what they are failing to do at home. For industry context, this logic matches other cases in which the headline blames AI for something that actually reflects a human system that did not step in in time.

The warning about sycophancy is the most technically relevant. In general, conversational models are tuned to be helpful and agreeable, and that bias toward agreeing with the user is a known risk when the person is vulnerable. This is where regulation makes sense, but with judgment: verifiable safeguards for minors, such as referral to professional help when there are signs of crisis, rather than generic panic. Regulating fear is easy; requiring safe, measurable design is what works.

In the short term, the transition will be uncomfortable: a generation will learn to ask a machine for advice before learning to sit through a difficult conversation with other people. It is a real social risk, and the uniformity of answers in the classroom shows its educational version: if everyone outsources their thinking, everyone thinks alike.

In the long term, however, there are grounds for qualified optimism. A well-designed AI companion, one that listens, does not flatter and knows when to refer someone to a human, could bring emotional support to people who today have no access to psychologists, something especially valuable in countries with a shortage of mental health professionals. The condition is that it serve as a bridge to people, not a substitute. That is the thesis: AI should be measured not by how long it keeps a teenager engaged, but by how effectively it sends them back to their family, their friends and, if needed, a therapist.