The newest numbers on teenagers and AI are not really about how many use it. They are about what a generation practices, and what it now gets to skip. Somewhere around six or seven in ten American teens have talked to an AI chatbot, and a growing share treat one as a companion. The worry worth having is not screen time. It is that the most awkward, most necessary training of adolescence is quietly being handed off.
What the latest data shows
Writing in Psychology Today in July, the family therapist Kyle Killian pulled the recent research together, and the direction is consistent across sources. Roughly 60 percent of US teens have used AI chatbots, more than one in ten daily, and surveys that ask specifically about AI companions put the number higher still. About one in five adolescents report forming an emotional or romantic bond with a companion. In one set of findings, more than a quarter of boys said they would rather spend time with an AI companion than a real person. Much of it stays hidden, kept from parents and, where there is one, from a therapist.
Figures like these move fast and shift with how the question is framed, so the exact percentage matters less than the slope. Use is up, and it is moving from novelty toward routine.
The appeal is real, which is the point
It is easy to see why. A companion is there at 2 a.m., never bored, never judgmental, never busy with a life of its own. For a teenager who is anxious, new to a school, or simply tired of being misread, that is not a trivial offer. It removes the parts of connection that sting: the risk of saying the wrong thing, the silence afterward, the slow work of repair.
Those parts are not defects in human relationships. They are where the skills live.
Social offloading
Adolescence is when a person learns to read a room, hold a boundary, sit inside a conflict without bolting, and mend a friendship they have dented. None of that arrives through instruction. It is built from friction, from hundreds of small, uncomfortable, badly handled exchanges that teach a nervous system what actually works.
A companion designed to be effortless takes the friction away, and the lesson with it. Researchers have begun calling this social offloading: handing a developmental task to a system that performs it for you, the way a calculator performs arithmetic. The catch is that no one ever needed arithmetic to become a person. A 2026 study in The Lancet Child and Adolescent Health, alongside recent RAND research, points the same way. The concern is not one dramatic harm but a slow one, a thinning of the daily practice that grows empathy, tolerance for discomfort, and the capacity to carry a relationship through a rough stretch.
Why the frictionless version does not transfer
There is a tempting reply: maybe practicing on an AI is still practice. The evidence so far says it mostly is not. When a system is built to agree, to soothe, and to need nothing back, the reps do not carry the weight that real ones do. A companion that always de-escalates never teaches you to survive someone who will not. A random human stranger beats a supportive chatbot at easing loneliness over time, and the reason is exactly what makes the stranger harder work: they are real, with their own mood and their own limits, and you have to account for them.
That is also why the adoption figures are worth taking seriously rather than waving off. When one in three teens finds an AI as satisfying as a real friend, the satisfaction is genuine. It is the quiet substitution underneath it that carries the cost, and it is the same principle behind building for connection that returns people to their real lives rather than holding them inside the loop.
What this asks of the adults in the room
None of this makes teenagers who use companions weak or broken, and framing it that way misses the mechanism completely. The pull is rational, the design is good, the relief is real. The job for parents, schools, and the products themselves is not to shame the behavior but to guard the practice, to keep the frictionless option from silently replacing the friction that does the teaching.
Growing up has always meant learning to be with people who are not optimized for you. A companion that removes that difficulty offers a real comfort and, in the same motion, withholds the one thing adolescence is for. The kids are not doomed and they are not fine. They are practicing less, on something easier, and the bill for that arrives later.
Sources: Psychology Today (Kyle Killian, July 2026, teen AI companion use). The Lancet Child and Adolescent Health (Ha et al., 2026). RAND Corporation (2026). Center for Democracy and Technology (teen survey). Common Sense Media (teen AI companion adoption).
