When a teen's first confidant is a chatbot: what 1 in 5 reveals

That one in five young people has turned to an AI chatbot to talk about how they feel is not a headline about technology, but about access. The figure was presented by Dr. Jessica Watrous, clinical director of Modern Health, on FOX 5's Good Day DC program on June 25, 2026. It is worth being honest about what we know and what we don't: the television segment does not detail the study's methodology —neither authorship, nor sample size, nor the age range that defines 'young'—, so the number deserves to be read as a snapshot of a trend, not as a closed statistic.

Even with that caution, the figure says something important. The adoption of AI as an emotional resource has ceased to be marginal. For a generation that grew up conversing with voice assistants and sharing their private lives on screens, writing to a system available at any hour, free of charge and free from the feeling of being judged is less intimidating than booking an appointment with a professional. That accessibility is real and, for those who would not otherwise seek help, it can be valuable.

The flip side is real too. A general-purpose chatbot is not designed to detect an acute crisis, does not monitor a patient's progress and can generate an emotional dependency that is difficult to measure. Here the boundary that matters is the one separating a tool integrated into a clinical circuit with human supervision —the model defended by a platform like Modern Health— from the loose use of a conversational assistant to talk about anxiety. For the clinician and the regulator that distinction is clear; for the young user it is often invisible.

Seen from the perspective of agentic AI, the terrain becomes even more delicate. A system that remembers someone's emotional history, identifies patterns of deterioration and proposes preventive interventions would have enormous potential, but it opens questions that no jurisdiction has fully resolved: who is liable if the system fails to detect suicidal ideation, or how the reasoning of an agent that operates continuously is audited. The European framework (EU AI Act) classifies AI systems aimed at diagnosis or treatment as high-risk, with requirements of transparency and human oversight, while in the United States the FDA is advancing guidance on AI medical software within a framework that is still evolving.

The constructive reading is not to block this entry point, but to design it well. If young people are already coming in this way, the urgent matter is that this first conversation be steered with judgment toward human help when needed, with clear limits on what a system can and cannot do. The FOX 5 DC figure is not the end of a debate, but proof that this debate is arriving late to a practice that is already an everyday reality.