Tens of millions of people now talk to AI about their feelings. A new study tries to answer the obvious question: what are they actually getting out of it — and what are they risking? The research, from Drexel University, finds that most users see AI mental-health tools as a supplement to human therapy, not a replacement.
What the study found
People turning to AI for mental-health support were typically seeking emotional support, empathy, reassurance for managing anxiety, coping strategies or simple companionship. The appeal is availability: an always-on, judgment-free listener at 2 a.m. costs nothing and never cancels. But the same users voiced clear concerns — about emotional dependence on a chatbot, exposure to misinformation, and overreliance on a tool that was never built for the job. The findings will be presented at the 2026 Annual Meeting of the Association for Computational Linguistics in San Diego in July.
The scale of the shift
This is no longer a fringe behavior. A 2025 American Psychological Association survey of U.S. adults found that almost half had used AI large-language-model tools for mental-health purposes in the past year, and more than 40 million people now use AI therapy apps every month. Demand on that scale, for something this sensitive, is exactly why researchers are paying attention.
The validation gap
The central worry is that people are leaning on general-purpose chatbots that were not designed or clinically validated for mental-health care. A model optimized to be agreeable can reinforce distorted thinking, miss warning signs, or offer confident but wrong guidance. Unlike a licensed therapist, it carries no duty of care and no accountability when it gets something seriously wrong.
Supplement, not substitute
The healthier framing the study points to is AI as a bridge: a place to vent, rehearse a difficult conversation or learn a coping technique between sessions, while the harder clinical work stays with a human professional. That distinction — companion versus clinician — is the one users themselves seem to be drawing.
The bottom line
AI mental-health tools are filling a real gap left by cost, stigma and therapist shortages, and people clearly find them useful. The task now is to keep them in their lane: helpful on the margins, transparent about their limits, and never mistaken for the care that only a trained human can provide.
Photo: Rawpixel Ltd / BY via flickr
