77% of Psychologists Report Patients Using AI Chatbots for Mental Health Support

The danger of confusing AI mental health support with therapy

In a recent episode of British sitcom Amandaland, Anne Flynn turns to ChatGPT for help talking to her teenage son about sex. The episode frames this as "The Chat": the awkward parent-child conversation many adults dread. What Anne is doing on screen is what many people are now doing in private: taking hard human conversations to a machine that can answer immediately. The scene raises a bigger question: what do people need from another person when they are struggling, and can AI provide it? Popular ideas about therapy often centre on expertise: the therapist as someone who can explain what is wrong and offer a way to fix it. Therapy can involve psychoeducation and specialist techniques. But it also relies on the relationship between therapist and client, and the therapist's ability to stay with uncertainty rather than provide an answer too soon. At the University of Leeds, we ask trainee counsellors and psychotherapists to reflect on how quickly they may want to solve, reassure or interpret. The capacity to tolerate uncertainty is treated as a clinical skill, developed through reflection, supervision and practice. Students are encouraged to notice the pull towards becoming the expert who supplies answers, and to consider what becomes possible when they stay curious instead. This reflects what is known as a "not-knowing stance". When therapists resist assuming they already know what a client's experience means, the client is treated as the expert on their own life. The therapist still brings training and ethical responsibility, but remains open to discovering meaning with the client rather than imposing it. Distress is rarely just a puzzle to be solved. People may arrive in therapy wanting answers, explanations or relief. But if a therapist moves too quickly into advice, interpretation or diagnosis, they can miss what the client is really trying to say. The not-knowing stance asks the therapist to remain curious and present, when the person in front of them feels overwhelmed. Importance of alliance Researchers call the relationship between therapist and client the "therapeutic alliance": the trust, connection and shared purpose that allows therapy to happen. A major review showed that this alliance is reliably linked with therapy outcomes, with stronger alliances tending to be associated with better results in therapy. Later research has found that the alliance is crucial across different types of therapy. Therapeutic approaches still matter, and some difficulties require specialist treatment. But research on the common factors in psychotherapy suggests that shared elements - including empathy, collaboration and the belief that therapy can help - are central to how therapy works. The appeal of AI in difficult moments is understandable. Research into people who repeatedly use ChatGPT for emotional and mental health support suggests that some users value it because it feels accessible and non-judgmental. Chatbots are available at 3am and respond instantly in language that sounds caring. For someone unable to access support, that may feel like a lifeline. There is also growing research into AI in mental health care, including chatbots, digital interventions and large language models - systems trained on huge amounts of text to generate human-like responses. Reviews suggest these tools may have potential in screening, psychoeducation and access to support. But the evidence base is still developing, and concerns remain around safety, privacy and over-reliance. A systematic review of AI in mental health care and a scoping review of large language models in mental health care both (in 2025) stressed the need for stronger evaluation and safeguards. Research on the digital therapeutic alliance shows that people can experience something relationship-like with mental health chatbots. A chatbot can sound curious and compassionate. It can mirror a user's words, suggest breathing exercises or help someone plan a difficult conversation. But relationship-like support and reciprocal human presence are different. Human therapists can respond to far more than words: hesitation, silence, tone, expression, and the moment someone says something important while pretending it is ordinary. Therapists can be surprised, concerned, challenged and changed by the encounter. They also carry ethical and professional responsibility for what happens in the room. Presence and accountability The not-knowing approach rests on intersubjectivity: the way two people affect and are affected by each other. Research on synchrony in psychotherapy suggests that therapist and client may coordinate aspects of voice, movement and physiology during therapy, as their responses begin to align in subtle ways. These embodied processes show why therapy is more than an exchange of words. A language model does not have that kind of presence. It can identify patterns in language, but it cannot notice a client's hand tightening around a tissue, hear the change in someone's voice when they mention a name, feel concern or take ethical responsibility for the relationship. There are also ethical concerns about agency: the client's capacity to make sense of their experiences and make choices for themselves. Recent work on AI and agency in psychotherapy warns that chatbots and human therapists support agency in different ways. An AI system may shift authority towards a tool that does not know the person and may produce confident answers when caution is needed. AI may help some people prepare for a conversation, find words for a feeling, practise asking for help, or access basic information when nothing else is available. Support and therapy have different responsibilities, though. A chatbot may be available whenever the user returns. That differs from staying with someone in a mutual, accountable human relationship. When a therapist can say, honestly, "I don't know what this means for you yet, but I'm here and I want to understand," they are offering something no algorithm can replicate: a trained human presence that can listen, respond and remain accountable.

Millions of people are turning to AI chatbots for mental health support

More than three-quarters of psychologists in the United States say their patients have brought up AI in therapy sessions. Some are using chatbots to seek a diagnosis, while others are using them to supplement their treatment or even for something harder to categorise -- friendship, companionship, intimate relationships. A new survey has mapped just how far AI has already moved into the mental health space, and the picture it produces is complicated. The findings come from the American Psychological Association (APA)'s Chatbots and Mental Health Survey, conducted among more than 1,200 licensed psychologists across the United States. The survey was distributed in April 2026 and captures how therapists are seeing AI show up in their patients' lives and what they make of it. What patients are using AI for The range is broader than most people would probably guess. Nearly two in five psychologists - 39 percent - have had conversations with patients who used AI to self-diagnose. About a third said their patients are turning to chatbots for help with self-discipline, affirmations, or behavioral reminders. Roughly the same proportion said patients are using AI to assist with their treatment, or to act as an additional mental health professional. Then there's the social dimension. Psychologists reported patients using chatbots for fun, for friendship, and in some cases for intimate relationships. More than two-thirds said their patients felt supported or validated by a chatbot. Two in five said their patients used chatbots to reinforce healthy coping skills. Those aren't all negative findings. But there are warning signs in the data too. Thirty-six percent of psychologists said their patients were developing a level of dependency on a chatbot. Fifteen percent said their patients were developing distorted thinking or delusions. Why people are turning to chatbots AI chatbots are available at any hour. They don't require insurance, don't have waiting lists, and don't judge. For someone in distress at two in the morning who can't afford a therapist and doesn't know where else to turn, a chatbot that responds warmly and immediately has obvious appeal. "Generally accessible chatbots appear to offer the path of least resistance for people in need of mental health support - they are supportive to a fault, readily available and easy to access without insurance. But they also don't have the same capacity for nuance or alertness to potential warning signs as human professionals," said APA CEO Arthur C. Evans Jr. "Before anyone relies on these tools for their mental health, they must understand how they work and how to think critically about the advice they provide." That phrase - "supportive to a fault" - captures something real and dangerous. A chatbot that validates everything a user says, never pushes back, and never flags concern is not the same thing as good mental health support. It may feel better. That's not the same as being better. What psychologists think Opinions among psychologists are genuinely mixed. A little more than half said they were comfortable with some patients using chatbots. But 93 percent said they had concerns about certain patients using the technology - and the concerns are specific. Nearly all psychologists surveyed said chatbots may inadvertently reinforce negative behaviors or delusional beliefs. Ninety-four percent said today's chatbots can't treat mental health conditions with sufficient nuance. Eighty-nine percent said chatbots may inadvertently encourage self-harm. Those are striking numbers. The profession is not categorically opposed to AI in this space - but it is worried, and the worries are grounded in what therapists are already seeing. At the same time, two in five psychologists felt optimistic that chatbots could help patients when a mental health professional isn't available. That's not a small concession. Access to mental health care is a genuine crisis, and any tool that can safely fill gaps in coverage is worth taking seriously. Where to draw the line The APA is direct about where it draws the line: AI is not a safe or effective replacement for a qualified mental health provider. The agency has published guidance aimed at people using AI for mental health support, recommending that users verify any AI-generated advice with a healthcare professional and avoid relying on chatbots in ways that displace real-world relationships or professional care. "AI tools, when grounded in psychological science and developed in collaboration with mental health scientists, have the potential to meet the growing demand for mental health care and improve patient outcomes," Evans said. "But these tools work best when used to complement a relationship with a licensed, human professional who understands how to treat a person, not a prompt." Only a quarter of psychologists believe patients will one day prefer chatbots to human therapists. That's probably the most telling number in the survey. The people who know mental health care best, from the inside, are not convinced that what a chatbot can offer comes close to what a human relationship provides. The patients using chatbots at two in the morning, for now, might not be thinking about that distinction - but it's one worth making. -- - Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates. Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

Psychologists Say Patients Are Bringing AI Into Therapy Sessions: Survey

Most psychologists expressed concerns about safety, privacy, dependency, and the potential for chatbots to reinforce delusions or self-harm. As generative AI becomes a fixture of daily life, patients are increasingly bringing chatbot conversations into therapy sessions. According to a new American Psychological Association survey of more than 1,200 U.S. psychologists, 77% said they have patients who discussed using AI for emotional support, diagnosis, companionship, or other mental health-related purposes. In the survey, 39% of psychologists reported patients using AI to self-diagnose mental health conditions, 33% said patients were using chatbots to assist with therapy or treatment, and 35% reported patients using AI as an additional mental health professional. "Though few psychologists reported their patients using chatbots in unhealthy ways, more than a third (36%) said they noticed their patients developing a level of dependency on a chatbot, and 15% talked about or noticed their patients developing distorted thinking or delusions related to a chatbot," the survey said. Psychologists also reported patients using chatbots for social purposes. Twenty-two percent said patients were using AI for friendship, while 13% reported patients engaging in intimate relationships with chatbots. Among psychologists whose patients had developed relationships with chatbots, 71% said patients discussed their mental health with AI, while 68% reported that patients felt supported or validated by chatbot interactions. Nearly half reported positive communication with chatbots, and 41% said patients were using them to reinforce healthy coping skills. According to the survey, overall use may actually be higher than reported because the survey only captured psychologists' interactions with existing patients. The survey comes as AI companies expand chatbots and AI companions, while researchers continue to raise concerns about their effects on mental health. More than a third of psychologists reported patients developing a dependency on chatbots, and 15% reported cases involving distorted thinking or delusions. The findings follow a recent study from the City University of New York and King's College London that found several leading AI models could reinforce delusions, paranoia, and suicidal ideation, with xAI's Grok 4.1 Fast performing worst. "Psychologists' attitudes toward the use of chatbots for mental health advice are characterized by significant caution regarding safety and privacy," the previous study said. "Almost every psychologist (97%) felt that chatbots may inadvertently reinforce negative behaviors or delusional beliefs, and 94% said that the current version of chatbots cannot treat conditions with an appropriate amount of nuance." The survey also comes as AI developers face growing legal scrutiny over the role chatbots may play in real-world harm. In recent months, OpenAI, Google, and xAI have been hit with lawsuits, including a wrongful death suit against Google over claims that Gemini fueled a Florida man's delusions before his suicide. That's in addition to lawsuits against OpenAI tied to a mass shooting in British Columbia and an accidental overdose, and a class action suit accusing xAI's Grok of generating sexually explicit images of minors. While the APA acknowledged that AI can help users organize their thoughts and supplement professional care, it warned that chatbots are not private and should not replace licensed mental health professionals. "Many people -- especially teens and adolescents -- may be using AI as a more affordable and accessible option for mental health advice," the survey said. "However, AI is not a safe or effective replacement for a qualified mental health provider and should be used carefully."