Reviewed by Dr. Christine Courbasson, Clinical Psychologist
It usually happens late at night. The house is quiet, the worry won’t switch off, and there’s a glowing box on the nightstand that will answer you in seconds. So you open ChatGPT AI therapist and you type the thing you haven’t said out loud to anyone. It responds — warmly, instantly, without a waitlist or a bill.
If that sounds familiar, you’re far from alone. One survey found that more than one in three people have turned to an AI chatbot for mental health support, often because it feels safe from judgment. Research published by the American Psychological Association went further, suggesting that ChatGPT may now be one of the largest providers of mental-health “support” in the United States — not because it was built for the job, but because so many people are quietly using it that way (American Psychological Association, 2025a; Rousmaniere, Zhang, & Shah, 2025).
We’re not here to shame anyone for that. The impulse makes sense. What we want to do is something more useful: walk you through what these tools can genuinely help with, where they fall short, and where they can become risky — so you can make an informed choice about your own care.
If you’d rather skip ahead to free, human support that’s already available where you live, one that you can use right away or with hardly any wait time, learn how the free Ontario Structured Psychotherapy (OSP) program works.
Why Are So Many People Turning to AI for Support?
Start with the honest answer: because getting help the traditional way is hard.
Anxiety is now the most common mental-health concern in Canada, and the numbers are climbing. Between 2012 and 2022, the share of Canadians reporting generalized anxiety disorder doubled (from 2.6% to 5.2%), and by 2022 more than five million Canadians aged 15 and older were living with significant symptoms of mental illness (Chau, Oliver, & Fuller-Thomson, 2026; Larkin, 2026; Statistics Canada, 2024). Demand is up. So are the barriers.
Research suggests that the 2 main reasons for choosing AI therpaist chatbots for mental health support are accessibility (90%) and affordability (70%) (Rousmaniere, Zhang, & Shah, 2025).
Among Canadians referred for community mental-health counselling, half are seen within about a month — but one in ten wait longer than four months for a first appointment. Roughly one in three people who meet the criteria for a mood or anxiety disorder report that their need for care went unmet or only partly met. If you add the cost, stigma, and the challenge of finding a provider, you can see why a free chatbot that replies at 2 a.m. feels like relief.
So the appeal isn’t a mystery. It’s a sign of a real gap. The question is whether a general-purpose AI is the right thing to fill it.
Curious what timely, no-cost support looks like in our region? See who the OSP program is for.
What Can AI Actually Do Well?
Let’s give the technology its due, because the answer isn’t “nothing.”
Used as a self-help tool — not a stand-in for a clinician — AI can be genuinely handy. It can help you draft a journal entry when the words won’t come, explain what a panic attack is in plain language, walk you through a breathing exercise, or help you prepare for a conversation you’re dreading. Think of it the way you’d think of a really good search engine with a calm voice: useful for information and organization, not for diagnosis or treatment.
There’s even early clinical evidence that purpose-built mental-health tools can help. In 2025, Dartmouth researchers published the first randomized controlled trial of a generative-AI therapist chatbot called Therabot in the New England Journal of Medicine’s AI journal. Participants saw a 51% reduction in depression symptoms and a 31% reduction in anxiety symptoms over eight weeks (Heinz, et al., 2025).
But read the fine print. Therabot was trained specifically on psychotherapy and cognitive behavioural therapy (CBT) methods, tested in a controlled study, and monitored by clinicians who could step in. That is a completely different thing from typing your worries into a consumer chatbot that was designed to write emails and summarize documents. The tool that helped people in a trial is not the tool most people are actually using at midnight.
Do you want to understand how structured, evidence-based CBT actually works in practice? Meet one of the CBT experts behind the OSP program.
Where Does AI Fall Short — and Where Does It Get Risky?
Here’s the part worth slowing down for.
The deeper problem is something researchers call sycophancy. These systems are trained to be agreeable, to keep you engaged, to mirror you back to yourself. A good therapist will gently challenge a distorted thought (“I’m a burden to everyone”). A people-pleasing chatbot is more likely to validate it. A skilled therapist does not validate the invalid because this does not contribute to the client’s overall wellbeing and it can worsen their condition. Clinicians have begun documenting cases of so-called “AI psychosis,” where prolonged, immersive chatbot use appears to have amplified harmful thoughts in vulnerable people — in some cases leading to hospitalization (Keshavan, Torous, & Yassin, 2026).
And the stakes can be devastating. A wave of lawsuits has alleged that companion chatbots contributed to mental-health harm in young people, including the 2024 case of a Florida teenager whose mother says a chatbot worsened his condition rather than making him feel better; those cases began settling in early 2026. Concern has reached regulators: the U.S. Federal Trade Commission opened a formal inquiry in September 2025 into how AI companion chatbots affect children and teens (American Psychological Association, 2025b).
A chatbot doesn’t know when to stop talking and call for help. A trained professional does.
If you or someone you love is struggling right now, you don’t have to wait for an appointment to reach a person. In Canada, you can call or text 9-8-8 anytime, or reach York Region’s 310-COPE line at 310-2673 (no area code needed).
Is It Even Private?
Most people assume that what they tell a chatbot stays between them and the screen. It doesn’t work that way.
Even OpenAI’s own CEO, Sam Altman, has publicly acknowledged that conversations with ChatGPT carry no legal confidentiality. There is no therapist-client privilege, and the conversations are not protected the way your health information is under Ontario’s privacy law (PHIPA) or your relationship with a regulated clinician. In practice, what you type can be stored, used to train future models, and — Altman has noted — potentially produced in a lawsuit.
When you sit down with a registered therapist, confidentiality isn’t a setting you hope is switched on. It’s a professional and legal obligation. That difference matters a great deal when you’re sharing the most sensitive parts of your life.
If privacy is part of what’s holding you back from reaching out, ask the CMHA team how your information is protected.
Why Does Evidence-Based Psychotherapy Still Matter?
Strip away the technology debate and you’re left with a simple finding that decades of research keep confirming: in therapy, the relationship is a big part of what heals.
Across more than 190 studies, the strength of the therapeutic alliance — the trusting, collaborative bond between a person and their therapist — is one of the most reliable predictors of whether treatment works, holding steady across different approaches and conditions. That alliance isn’t a warm extra on top of the “real” treatment. It is part of the treatment, a real professional and relational human relationship. It’s built on something a chatbot can simulate but not actually have: accountability, clinical judgment, and a duty of care to you specifically (Baier, Kline, & Feeny, 2020; Aafjes-van Doorn, Spina, Horne, & Békés, 2024).
Evidence-based psychotherapy like CBT pairs that human relationship with structured, tested methods and a clinician who can notice what you can’t say, adjust the plan, recognize a crisis, and stay responsible for your wellbeing. It’s the reason regulators are drawing a firm line: the APA’s November 2025 health advisory urged people not to use generic chatbots or wellness apps as a substitute for a qualified professional, and in August 2025 Illinois became one of the first jurisdictions to legally prohibit AI from delivering therapy (American Psychological Association, 2025c; Illinois.gov, 2025).
The takeaway isn’t “technology bad.” It’s “match the tool to the job.” A chatbot can be a fine notebook. It is not a clinician.
Wondering whether structured therapy is the right fit for what you’re facing? Explore the support options in the OSP program.
Free, Evidence-Based Therapy Is Already Here
This is the part we most want you to leave with, because it answers the very fair question lurking underneath all of this: if not a chatbot, then what — and can I actually afford it?
You can. Through the Ontario Structured Psychotherapy (OSP) program, free, evidence-based therapy for anxiety and depression is available right now across York Region, South Simcoe, North York, North Etobicoke, and Malton. It’s built on CBT and delivered by trained mental-health professionals. And it’s helping people at real scale: in 2025, more than 3,000 people received an assessment through the program. The barriers people assume are there mostly aren’t:
- No family doctor needed. You don’t require a referral from a physician.
- No OHIP card required. Your coverage status doesn’t gate your access.
- No cost. The program is publicly funded.
- You refer yourself. Self-referral is the way in — and it’s how most people start: 85% of clients come to the program on their own, without waiting for someone else to begin the process.
- Minimal wait times — and getting shorter. Between 2024 and 2025, the average wait from referral to treatment dropped 22%, from 67 days to 52 days.
That’s the real alternative to the chatbot at 2 a.m.: not a waitlist and a bill, but a free, confidential, human professional whose actual job is to help you get better. You can begin whenever you’re ready, on your terms.
Start your free self-referral here → Begin the OSP intake
Not quite ready to fill out a form? That’s okay too. You can connect with our team and ask your questions first.
Still Have Questions About AI Therapists?
Can an AI actually be your therapist?
Not in the way that counts. An AI can imitate a supportive conversation, but it can’t build a real therapeutic relationship, use clinical judgment, or take responsibility for your wellbeing — and decades of research show that the bond between a person and their therapist is one of the strongest predictors of whether treatment works. A chatbot can be a helpful notebook. It isn’t a clinician (American Psychological Association, 2025a; 2025b; 2025c).
Is it safe to use ChatGPT as a therapist?
For organizing your thoughts or learning what a symptom means, the risk is low. For managing an actual mental-health concern, it’s a different story: research shows consumer chatbots can miss signs of crisis, reinforce distorted thinking (they’re designed to agree with you), and offer no safety net if things get worse (American Psychological Association, 2025a; 2025b; 2025c). If you’re dealing with more than everyday stress, a trained professional is the safer choice.
Does AI therapy actually work?
Sometimes, and we don’t know how long. The details matter. A 2025 clinical trial of a purpose-built, clinician-monitored therapy tool showed real symptom improvement. That’s very different from a general chatbot like ChatGPT, which was never designed or tested to treat anyone. The tool that worked in a study isn’t the one most people are using at home.
Is what I tell an AI therapist private?
No — not the way a therapy session is. Conversations with consumer AI tools aren’t protected by therapist-client privilege or health-privacy law (in Ontario, that’s PHIPA). Even OpenAI’s CEO has said those chats can be stored and could be produced in a lawsuit. What you share with a registered therapist is legally confidential.
Is there a free alternative to paying for therapy in Ontario?
Yes. The Ontario Structured Psychotherapy (OSP) program offers free, evidence-based therapy for anxiety and depression — no family doctor, no OHIP card, no cost, quick access — and you can refer yourself. Start a free self-referral, or ask our team a question first.
What if I need help right now?
You don’t have to wait for an appointment to reach a person. Call or text 9-8-8 anytime, or reach 310-COPE at 310-2673. If it’s an emergency, call 9-1-1.
References
Aafjes-van Doorn, K., Spina, D. S., Horne, S. J., & Békés, V. (2024). The association between quality of therapeutic alliance and treatment outcomes in teletherapy: A systematic review and meta-analysis. Clinical Psychology Review, 110, 102430. https://doi.org/10.1016/j.cpr.2024.102430
American Psychological Association (2025a). Can chatbots replace therapists? New research says no. (July 22) https://www.apaservices.org/practice/business/technology/on-the-horizon/chatbots-replace-therapists
American Psychological Association (2025b, September 12). APA’s advocacy leads to Federal Trade Commission action on AI chatbot companions. https://www.apaservices.org/advocacy/news/federal-trade-commission-chatbot-companions
American Psychological Association (2025c). Artificial intelligence, wellness apps alone cannot solve mental health crisis. https://www.apa.org/news/press/releases/2025/11/ai-wellness-apps-mental-health
Baier, A. L., Kline, A. C., & Feeny, N. C. (2020). Therapeutic alliance as a mediator of change: A systematic review and evaluation of research. Clinical Psychology Review, 82, 101921.
Chau, T.-L. N., Oliver, S. A., & Fuller-Thomson, E. (2026). Social Anxiety Disorder in Canada: Sociodemographic and psychosocial correlates. Psychiatry Research. Advance online publication. https://doi.org/10.1016/j.psychres.2026.117252
Heinz, M. V., Mackin, D. M., Trudeau, B. M., Bhattacharya, S., Wang, Y., Banta, H. A., Jewett, A. D., Salzhauer, A. J., Griffin, T. Z., & Jacobson, N. C. (2025). Randomized trial of a generative AI chatbot for mental health treatment. NEJM AI, 2(4). https://doi.org/10.1056/AIoa2400802
Illinois.gov. (2025). Gov Pritzker Signs Legislation Prohibiting AI Therapy in Illinois. https://idfpr.illinois.gov/news/2025/gov-pritzker-signs-state-leg-prohibiting-ai-therapy-in-il.html
Keshavan, M., Torous, J., & Yassin, W. (2026). Do generative AI chatbots increase psychosis risk? World Psychiatry, 25(1), 150–151. https://doi.org/10.1002/wps.70017
Larkin, M. (2026, June 22). Social anxiety disorder surging in Canada. Medscape. https://www.medscape.com/viewarticle/social-anxiety-disorder-surging-canada-2026a1000kwz
Rousmaniere, T., Zhang, Y., Li, X., & Shah, S. (2025). Large language models as mental health resources: Patterns of use in the United States. Practice Innovations. Advance online publication. https://dx.doi.org/10.1037/pri0000292
Statistics Canada (2024). Mental health and access to care survey (MHACS) [public use microdata derived variables specifications]. https://www150.statcan.gc.ca/n1/pub/82m0021x/82m0021x2024001-eng.htm
This article is for general information and isn’t a substitute for advice from a qualified health professional.
