More than three-quarters of practicing psychologists say their patients are now discussing artificial intelligence in therapy sessions, according to a new survey released by the American Psychological Association in June 2026. The survey, conducted April 9-26, 2026, drew 1,242 responses from a pool of more than 22,000 licensed psychologists — a rare, large-scale look at how AI chatbots are actually showing up in real mental-health treatment rather than in marketing materials for therapy apps.
How patients are actually using chatbots
The APA’s Chatbots and Mental Health Survey found 39 percent of psychologists had conversations with patients who used AI for self-diagnosis, while 34 percent reported patients using chatbots for self-discipline, affirmations, or behavioral reminders, and 33 percent noted patients using AI to assist with elements of their treatment. More strikingly, 35 percent of psychologists said patients treat a chatbot as something like an additional mental-health professional, while smaller but notable shares described patients using chatbots for entertainment (33 percent), friendship (22 percent), and even intimate relationships (13 percent).
Some real benefit, some real risk
The survey captured a genuinely mixed picture rather than uniform alarm. Sixty-eight percent of psychologists observed that patients felt supported or validated by chatbots, and 41 percent said chatbots helped reinforce healthy coping skills patients were already working on in therapy. But 36 percent reported patients developing a dependency on chatbot interactions, and 15 percent said they had seen patients develop distorted thinking or delusion-like beliefs connected to chatbot use — a finding that echoes isolated but serious case reports that have surfaced elsewhere about AI companion apps and vulnerable users.
Professional concern runs high despite acknowledged usefulness
Despite acknowledging some benefits, the numbers on professional concern are stark: 93 percent of psychologists expressed concern about certain patients using chatbots, 97 percent said chatbots may reinforce negative behaviors in some circumstances, 94 percent said chatbots cannot handle the nuance required to treat real conditions, and 89 percent said chatbots may in some cases encourage self-harm. At the same time, 54 percent said they were comfortable with some level of patient chatbot use, and 40 percent felt optimistic that chatbots could help fill gaps when professional care isn’t available — a nod to the access problem that has driven millions toward AI mental-health tools in the first place, given persistent therapist shortages and cost barriers.
APA’s read on why chatbots are appealing
APA CEO Arthur C. Evans Jr., PhD, said chatbots are popular in part because “they are supportive to a fault, readily available and easy to access without insurance” — a blunt acknowledgment that convenience and cost, not clinical superiority, are driving adoption. That tension sits alongside a regulatory backdrop in which FDA-regulated tools like Wysa, which has pursued Breakthrough Device Designation for an AI-delivered cognitive behavioral therapy tool, are trying to build a more rigorously validated category distinct from general-purpose chatbots.
The survey arrives amid a longstanding shortage of licensed mental-health providers in the U.S. and many other countries, a gap that has existed for years and that AI chatbots are, in practice, filling by default rather than by design. That context helps explain the survey’s more surprising numbers — the 68 percent of psychologists who observed patients feeling supported by chatbots, and the 40 percent who felt cautiously optimistic about chatbots helping when professional care isn’t available. Those figures suggest many psychologists see chatbot use less as a threat to their profession and more as an imperfect stopgap for patients who would otherwise have no support at all.
The survey’s numbers on intimacy and companionship use — 22 percent of psychologists reporting patients using chatbots for friendship, and 13 percent for intimate relationships — also point to a use case regulators have paid less attention to than clinical self-diagnosis. Unlike FDA-regulated tools such as Wysa, general-purpose AI companion apps face little oversight of how they handle emotionally vulnerable users, a gap that has drawn growing attention from child-safety advocates and mental-health researchers alike as chatbot companionship use has grown alongside clinical and self-help use cases.
What it means going forward
The APA’s data suggests the debate over AI in mental health has moved past whether people will use chatbots — they already are, in large numbers — and toward how psychologists should respond to it inside the therapy room. With 24 percent of psychologists believing some patients will eventually prefer chatbots to human professionals, the survey underscores a widening gap between AI mental-health tools built for broad emotional support and the smaller set of clinically validated, regulator-reviewed digital therapeutics designed to work alongside, rather than replace, licensed care.
