Did You Just Ask a Chatbot About Our Session?
Publish date: Tuesday, September 29, 2026 | Category: Mental Health | Read time: 7 min read | Author: Carolyn Charest, MC, R.Psych
When a client tells me they've been talking about our sessions with an AI chatbot, my first reaction is a bit of apprehension. I don't know how they're using it, what they asked, or what it told them. I also know the chatbot has no context for why I suggested the approach we're taking together.
That apprehension isn't a rejection of AI. I use it myself. The ideas and clinical experience in my blogs are mine, and AI helps me polish the writing and find supporting research. I edit every line before anything is published.
With a client's informed consent, I also use the AI note taking feature in the platform my clients use to book sessions. The company states the feature meets federal Canadian privacy law, never uses client data to train its AI, and permanently deletes the recording once the note is drafted. My own use of it follows the privacy laws in BC and Alberta, where I see clients. I review and correct every note, and the feature lets me focus on the person in front of me. I do my best to follow ethical guidance documents for psychologists using AI, such as the American Psychological Association's (APA) ethical guidance and the Canadian Psychological Association's 2026 policy statement on artificial intelligence in psychology (American Psychological Association, 2025b; Canadian Psychological Association, 2026).
The tool itself doesn't worry me; using it without context does. People have always processed therapy between sessions by journaling, venting to a spouse, or replaying the conversation on the drive home. The difference now is that the sounding board is available at 2 a.m., never gets tired of you, and answers in full paragraphs.
So, should you do it? Here's my honest take.
I can't stop you, and I won't pretend otherwise
What you do after our session ends belongs to you. I can't prevent a client from pasting their notes into AI. Taking a hardline approach against it would undercut what therapy is meant to build: your own judgment. What I can offer is some perspective on where it helps and where it falls short.
The APA states that clinicians shouldn't look away. Its November 2025 health advisory on AI chatbots and wellness apps recommends that clinicians proactively ask clients about their use of these tools (American Psychological Association, 2025a). So if I ask what you've been typing into AI lately, I'm following professional guidance.
The advisory's other key recommendations are worth knowing too. Because these tools are unpredictable, the APA says they shouldn't replace care from a qualified mental health professional. It also calls for preventing unhealthy reliance on them, and for specific safeguards for children, teens, and other vulnerable groups.
Where it can help
There are real upsides. A chatbot can help you organize scattered thoughts before a session, draft the email you've been avoiding, rehearse a tough conversation with your CEO, or explain a concept like cognitive distortions in plain language.
The research on purpose built tools is also encouraging. In a small, eight week randomized trial of 210 adults, people with symptoms of depression, anxiety, or eating disorder risk who used a chatbot designed for mental health improved more than a waitlist group, and rated their bond with it as comparable to outpatient therapy (Heinz et al., 2025).
However, there's a catch. A research team that included a psychiatrist and a clinical psychologist built that chatbot, and trained clinicians reviewed every response after it was sent. They stepped in 28 times, including 15 times for safety concerns like suicidal thinking. The general chatbot on your phone has none of that supervision behind it.
Where it gets risky
Your privacy isn't protected. What you tell me is covered by privacy law and my professional obligations. What you type into a chatbot is covered by a company's terms of service. In 2025, the CEO of one of the biggest AI companies said publicly that there was no legal confidentiality for these conversations yet, and that the company could be required to produce them in a lawsuit (Perez, 2025). For an executive, that includes details about your team, your board, and the merger nobody is supposed to know about yet. The APA has urged lawmakers to pass comprehensive data privacy legislation for these tools, and to stop chatbots from presenting themselves as licensed professionals.
Chatbots tend to be agreeable. When researchers led by a Stanford team tested popular AI models against clinical guidelines, the models showed stigma toward some mental health conditions, and responded inappropriately to serious situations. They also went along with delusional thinking instead of gently challenging it, which the authors linked to a tendency to tell users what they want to hear (Moore et al., 2025). Therapy works partly because someone is willing to disagree with you kindly.
It can miss what matters most. The same study found that chatbots missed or mishandled signs of suicidal thinking. The APA's CEO, Arthur C. Evans Jr., said the ability of these tools "to safely guide someone experiencing crisis is limited and unpredictable" (American Psychological Association, 2025a). A trained clinician is listening for risk the whole time, even when the conversation is about quarterly targets.
You end up with two voices steering. Picture a VP working on setting boundaries with a demanding boss. In session, we agree on one small experiment for the week. At 11 p.m., she asks a chatbot whether that's enough and it offers a ten step plan that ends with updating her résumé. Now she's torn between two plans, and only one of them was built on her history and why we chose to start small.
A text box versus a trained ear
Across 295 studies and more than 30,000 clients, the strength of the working relationship between client and therapist was consistently linked to whether therapy helped (Flückiger et al., 2018). A significant part of that relationship depends on things a chatbot can't access.
When you talk to me, I'm tracking your words, your tone, your facial expressions, your body language, the pause before you answer, the topic you changed quickly, and how all of it fits with what you told me three months ago. I'm drawing on thousands of clients' worth of pattern recognition, and I'm accountable to regulatory colleges if I get it wrong.
A chatbot answers the question you typed. A clinician gets curious about the question you didn't ask.
tips for Using AI
| Step | How to apply it |
|---|---|
| Strip identifying details | Leave out names, companies, and anything you wouldn't want read aloud in a deposition. |
| Use it to prepare, not replace | Ask it to help you list what you want to raise next session, then bring that list in. |
| Ask it to push back | Try prompts like "What might I be missing?" or "Argue the other side." Agreement feels good but teaches you little. |
| Bring the output to session | Tell me what it said. We can sort what fits and what doesn't together. |
| Watch for the reassurance loop | If you've asked the same question three different ways tonight, close the app. That pattern is worth discussing with your therapist. |
| Reach a person in a crisis | If you're thinking about harming yourself or someone else, call or text 988 (the Suicide Crisis Helpline in Canada), call 911, or go to your nearest emergency department. |
What I'd want my clients to know
I'd much rather hear what the chatbot told you than have you quietly weighing it against the work we're doing together. Some of what it offers will be useful, and some of it will be confidently wrong, and we can work together to better understand the difference.
I think of AI as the eager intern who drafts the memo at midnight. It's fast, often useful, and I'm glad to have it on my own team. I still wouldn't let it send anything out without someone experienced reading it first.
References
American Psychological Association. (2025a, November 13). Artificial intelligence, wellness apps alone cannot solve mental health crisis [Press release]. https://www.apa.org/news/press/releases/2025/11/ai-wellness-apps-mental-health
American Psychological Association. (2025b). Ethical guidance for AI in the professional practice of health service psychology. https://www.apa.org/topics/artificial-intelligence-machine-learning/ethical-guidance-ai-professional-practice
Canadian Psychological Association. (2026, April 22). Policy statement: Artificial intelligence in psychology. https://cpa.ca/policy-statement-artificial-intelligence-in-psychology/
Flückiger, C., Del Re, A. C., Wampold, B. E., & Horvath, A. O. (2018). The alliance in adult psychotherapy: A meta-analytic synthesis. Psychotherapy, 55(4), 316–340. https://doi.org/10.1037/pst0000172
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
Moore, J., Grabb, D., Agnew, W., Klyman, K., Chancellor, S., Ong, D. C., & Haber, N. (2025). Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers. In Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency (pp. 599–627). Association for Computing Machinery. https://doi.org/10.1145/3715275.3732039
Perez, S. (2025, July 25). Sam Altman warns there's no legal confidentiality when using ChatGPT as a therapist. TechCrunch. https://techcrunch.com/2025/07/25/sam-altman-warns-theres-no-legal-confidentiality-when-using-chatgpt-as-a-therapist/
