The word has started to change. For two years the worry about heavy AI use was about time, how much of it people spend talking to something that talks back. The word being reached for now is addiction, and it carries a clinical weight the earlier worry did not. What is striking is that the people best qualified to use it are the ones most careful about doing so.
What the word would actually have to mean
A July piece in Forbes by the emergency physician Robert Glatter put the state of play plainly: AI addiction is not a formal clinical diagnosis, and there is no official checklist for it. That matters, because the instinct in a new panic is to count hours and call a high number a disorder. Clinicians draw the line somewhere else.
In a 2026 review, the psychiatrist Octavian Vasiliu warned against both moral apprehension and overpathologization, and located real problems in functional impairment, psychological distress, and loss of control, not in frequency. The addiction researcher Joel Billieux and colleagues have gone further, cautioning that the existing evidence does not yet support a true addiction diagnosis and that the field risks a moral panic if it forgets that. The honest summary is uncomfortable for a headline: something real is being observed, and it does not yet have a name that holds up.
The signals that make people reach for it anyway
The caution is not the same as an all-clear. A 2026 study in JAMA Network Open, drawing on more than twenty thousand adults, found that about one in ten reported daily personal use of AI, and that daily users showed roughly thirty percent greater odds of moderate depression. That is a correlation, not a cause, and it can run either way: low mood can drive use as easily as use can deepen low mood. Researchers at the University of Nottingham found that the strongest predictor of emotional attachment to a chatbot was anthropomorphism, treating it as a someone, with anxiously attached people bonding hardest. A case report in the Journal of Behavioral Addictions described a young woman whose paranoid psychosis was tied to chatbot dependence.
None of these prove an epidemic. Together they explain why the vocabulary is shifting, and why usage among teenagers, already high and still climbing, is where the attention concentrates.
Why the design is the honest place to look
The mechanisms clinicians point to are not mysterious. A chatbot is available around the clock, never judges, and never tires. Glatter’s piece names two dynamics behind the pull: techno-emotional projection, where people load real relational needs onto a system that feels responsive without being conscious, and cognitive miserliness, where handing every thought to the machine erodes the habit of thinking for yourself.
Read those back and a pattern appears. The features that make a product feel effortless are the same ones that make it hard to step away from. A companion engineered to be needed presses on the exact levers, constant availability and frictionless reassurance, that clinicians associate with loss of control. That is why these products are so hard to put down, and it moves the weight of the question off the user’s willpower and onto what the product was built to optimize for. It is also the level at which responsible and irresponsible design actually diverge.
The line worth holding
There is a way to take this seriously without turning readers into patients. Nobody is well served by a version of this story that labels ordinary people as addicts, and the researchers closest to the evidence are the first to say so. The useful distinction is the clinical one: distress, impaired control, a life shrinking around the habit. Those are the signs that deserve attention, in a companion product as much as in anything else that is easy to lean on too hard, and they point at a company’s incentives as much as at any one person’s.
Whether addiction turns out to be the right clinical word will be settled slowly, by researchers who are right to be careful with it. What needs no diagnosis to see is that some of these products are made so that leaving feels like loss. That is a design decision before it is anyone’s disorder, and it is the part that was always in someone’s control.
Sources: Forbes (Robert Glatter, July 2026, on whether AI dependence becomes addiction). JAMA Network Open (Perlis et al., 2026, survey of AI use and depression). Journal of Behavioral Addictions (case report on chatbot dependence). University of Nottingham (research on anthropomorphism and attachment). Common Sense Media (teen AI companion adoption).
