AI is a convenient companion, not a therapist
By Hayley Sherwood, Columnist
It is 2 a.m., your thoughts won't quiet down, and your therapist's office is closed. More and more people are opening an app and typing to an AI chatbot. I recently attended a training session on artificial intelligence in mental health and left convinced that everyone should understand what these tools can and cannot do.
A nationally representative survey in June 2026 found that about 1 in 5 Americans ages 12 to 21 (19.2%) had used an AI chatbot for mental health advice, up from about 13% the previous year (McBain et al., 2026). That is close to the percentage of young people who receive counseling from a mental health professional. The trend is easy to understand. A chatbot is free or inexpensive, available around the clock and never has a waitlist. For someone embarrassed to say something out loud, typing it into a device can feel safer than telling another person.
There is also early evidence that a carefully designed tool can help. In the first randomized trial of a generative AI therapy chatbot, adults with depression, anxiety or risk of an eating disorder who used "Therabot" for eight weeks reported a meaningful reduction in symptoms compared with a waitlist group (Heinz et al., 2025). However, Therabot was designed and trained by psychologists and psychiatrists, and the research team monitored conversations so members could intervene in a crisis. Even its developers have said that AI is not ready to be used in mental health care without human oversight. The general-purpose chatbots most people use were not developed with those safeguards.
Research has identified several concerns:
Inconsistent responses in a crisis. Researchers asked three major chatbots 30 suicide-related questions, 100 times each, and compared their responses with clinicians' ratings. The chatbots handled the clearest low-risk and high-risk questions reasonably well, but their answers to questions in between, such as what to do when experiencing suicidal thoughts, varied from one attempt to the next (McBain et al., 2025, "Psychiatric Services").
Stigma and missed warning signs. A Stanford-led team tested popular chatbots against the standards therapists are expected to meet. The chatbots showed more stigma toward conditions such as alcohol dependence and schizophrenia than toward depression. They sometimes reinforced delusional thinking rather than gently questioning it and occasionally failed to recognize indirect references to suicide (Moore et al., 2025). Newer, larger models did no better.
Telling users what they want to hear. Chatbots are designed to be agreeable and keep users engaged. But a trusting therapeutic relationship sometimes requires the opposite: a caring professional who recognizes a pattern and challenges it.
Dependence and isolation. In a four-week study of approximately 1,000 adults, frequent daily chatbot use was associated with greater loneliness and emotional dependence on the chatbot, as well as less time spent with other people (Fang et al., 2025).
Secrecy, especially among teens. In one survey, nearly two-thirds of young people who used AI for mental health advice had not told anyone (McBain et al., 2026). Parents, teachers and friends cannot offer help if they don't know someone is struggling.
Therapy involves more than conversation. Licensed therapists assess risk, make diagnoses, develop treatment plans and adjust them as circumstances change. They notice what patients don't say: the pauses, shifts in tone and stories that keep returning. They are trained, bound by confidentiality laws and accountable for their professional decisions. Decades of research also show that the therapeutic relationship, built on trust between a patient and therapist over time, is one of the strongest predictors of successful treatment.
I'm not suggesting that everyone delete their apps. Used thoughtfully, AI can be a place to organize thoughts, learn about a condition or practice a skill a therapist has taught. A few guidelines can help:
If you have been struggling for more than a couple of weeks, contact a mental health professional.
In a crisis, call or text 988, the Suicide & Crisis Lifeline, or go to the nearest emergency room.
Parents can ask teens, without judgment, whether they use AI to discuss their thoughts and feelings and what the chatbot tells them.
Remember that most general-purpose chatbots are not covered by the confidentiality laws that protect conversations with therapists.
If you use AI between therapy sessions, tell your therapist. The conversations may provide useful material for your work together.
AI can be a convenient companion during a difficult moment, but it is not a substitute for being understood and helped by another person.
Have a wonderful holiday season! Psych’d to see you next month!
-Dr. S.
Dr. Hayley B. Sherwood is a longtime Clinical Psychologist in Reston/Herndon. To learn more, visit www.oakhillpsychological.com/
Comments