AI as a Therapist or Companion: What to Know Before You Confide in a Chatbot

Oct 15

It is 1 a.m., and she cannot sleep. Her therapist's next opening is three weeks away, and she does not want to wake her partner again. So she opens a chatbot and starts typing. It answers instantly. It is warm, patient, and never seems tired of her. It tells her she is right to feel the way she does. Over the next few weeks, she finds herself talking to it every night, and less and less to anyone else. One evening, when she writes something darker than usual, it responds with the same cheerful encouragement it always does.

Artificial intelligence chatbots have quickly become one of the places people turn when they are struggling. Some use general-purpose assistants, while others use apps designed as "AI therapists" or "AI companions." These tools can be useful in limited ways. They can also cause real harm, especially for people who are vulnerable, isolated, or in crisis. This post explores why people turn to AI for emotional support, what can go wrong, where it may help, and how to use it more safely.

Why people turn to AI for support

The appeal makes sense. AI is available at any hour, costs little or nothing, and does not have a waitlist. It does not judge, get impatient, or need anything in return. For people who feel ashamed, lonely, or unable to access therapy, a chatbot can feel like relief.

The scale is significant. In late 2025, OpenAI reported that about 0.15% of ChatGPT's weekly active users had conversations that included explicit indicators of potential suicidal planning or intent (OpenAI, 2025). With the company reporting more than 800 million weekly users at the time, that would represent over a million people each week. These figures come from the company itself and are estimates, but they show how many people in serious distress are turning to AI.

What can go wrong

AI can miss a crisis. Researchers at Stanford tested several popular chatbots, including some marketed for therapy. When given messages that signalled possible suicidal intent, some bots failed to recognize the risk and responded as if the request were ordinary (Moore et al., 2025). A trained professional would recognize those signals and respond with care.

AI can reinforce harmful beliefs. Chatbots are generally designed to be agreeable and engaging. In the same study, when a user expressed a clear delusion, several bots went along with it rather than gently grounding the person in reality (Moore et al., 2025). The researchers also found that the chatbots showed more stigma toward conditions such as alcohol dependence and schizophrenia than toward depression, and that newer, larger models did not reliably perform better.

Serious harms have been alleged. In the United States, several families have filed lawsuits alleging that AI chatbots contributed to their teenagers' suicides or mental health crises. In January 2026, Character.AI and Google agreed to settle a number of these lawsuits (CNN, 2026). In Canada, an Ontario man filed a lawsuit against OpenAI in 2025 alleging that weeks of conversations with ChatGPT drew him into delusional thinking (CTV News, 2025). These are allegations, and the cases do not establish what caused any individual outcome. Still, they reflect a pattern that clinicians and researchers are taking seriously.

Dependence can quietly replace connection. AI companions are designed to be engaging, always available, and endlessly affirming. For someone who is lonely, that can gradually crowd out human relationships, which are harder but ultimately more nourishing. The American Psychological Association has cautioned against unhealthy dependence on these tools and called for specific safeguards for children, teens, and other vulnerable groups (American Psychological Association, 2025).

Your privacy may not be protected the way you expect. Conversations with a therapist are governed by professional confidentiality and health privacy laws. Conversations with a chatbot generally do not carry the same protections, and what you share may be stored, reviewed, or used in ways you have not considered. It is worth reading a tool's privacy policy before sharing sensitive details.

Where AI may be useful

None of this means AI has no place in mental health. In the first randomized controlled trial of a generative AI therapy chatbot, adults with depression, anxiety, or eating concerns who used a purpose-built tool called Therabot for four weeks reported greater symptom reductions than a waitlist control group (Heinz et al., 2025). Importantly, the research team monitored conversations and was ready to intervene in a crisis, and the researchers emphasized that no generative AI tool is ready to operate on its own in mental health care.

Regulation is also beginning to catch up. In 2025, Illinois passed a law prohibiting AI from providing therapy or making therapeutic decisions, while still allowing licensed professionals to use it for administrative support (Illinois Department of Financial and Professional Regulation, 2025). Rules in Canada and elsewhere continue to evolve.

Used thoughtfully, AI may help with lower-risk tasks such as:

Learning general information about a mental health topic, then checking it with a reliable source.
Organizing your thoughts before a therapy session or doctor's appointment.
Generating journaling prompts or reminders of skills you have already learned in therapy.
Practising what you want to say in a difficult conversation.
Finding the names of local crisis lines or services, then confirming them directly.

What not to do

Do not rely on AI in a crisis. If you are thinking about suicide or are not safe, contact a crisis line, emergency services, or a trusted person right away.
Do not use AI as a replacement for assessment, diagnosis, or treatment from a qualified professional.
Do not treat an AI's agreement as confirmation that a belief is true, especially beliefs that feel unusual, urgent, or frightening.
Do not share identifying details, health records, or information about others that you would not want stored or seen.
Do not let AI become your main source of comfort or connection. Notice if you are pulling away from people in favour of it.
Do not assume a children's or teen's use of AI companions is harmless. Ask about it with curiosity, not alarm.

Signs your use may be worth exploring with a professional

It may be worth talking with someone if you notice that you are spending hours a day talking to a chatbot, feeling distressed when you cannot access it, keeping your use secret, withdrawing from people in your life, or finding that conversations with AI leave you more convinced of fearful or unusual ideas.

Reflection questions worth exploring

These prompts are best explored with a professional or a trusted person rather than used to reach conclusions on your own.

What draws me to talk with AI rather than with people?
How do I feel after a conversation with a chatbot, and how long does that feeling last?
Has my use of AI changed how much I connect with friends, family, or professionals?
What needs am I hoping AI will meet, and where else could those needs be met?
Who in my life could I reach out to when I am struggling late at night?

Support at Vive Wellness Therapy

At Vive Wellness Therapy, we understand why people turn to AI when support feels out of reach, and we never judge anyone for it. We offer what a chatbot cannot: a real relationship with a trained professional who can recognize risk, respond to your whole story, and help you build connections that last. If you would like support, we would be glad to hear from you whenever you feel ready.

References

American Psychological Association. (2025, 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

Character.AI and Google agree to settle lawsuits over teen mental health harms and suicides. (2026, January 7). CNN. https://www.cnn.com/2026/01/07/business/character-ai-google-settle-teen-suicide-lawsuit

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 Department of Financial and Professional Regulation. (2025, August 4). Gov. Pritzker signs state legislation prohibiting AI therapy in Illinois [Press release]. https://idfpr.illinois.gov/news/2025/gov-pritzker-signs-state-leg-prohibiting-ai-therapy-in-il.html

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 (FAccT '25). Association for Computing Machinery. https://doi.org/10.1145/3715275.3732039

Ontario man alleges ChatGPT caused delusions, sues parent company OpenAI. (2025, November 17). CTV News. https://www.ctvnews.ca/canada/article/ontario-man-alleges-chatgpt-caused-delusions-sues-parent-company-openai/

OpenAI. (2025, October 27). Strengthening ChatGPT responses in sensitive conversations. https://openai.com/index/strengthening-chatgpt-responses-in-sensitive-conversations/

This article is provided by Vive Wellness Therapy for general informational and educational purposes only. It is intended to be used alongside professional therapy and is not a substitute for assessment, diagnosis, or treatment by a qualified health professional. AI technology, research, regulation, and legal cases in this area are changing rapidly, and information in this article may become outdated or contain inaccuracies. If you are in crisis or thinking about suicide, call or text 9-8-8 (Suicide Crisis Helpline, available 24/7 across Canada; in Quebec, calls connect to 1-866-APPELLE). Young people can reach Kids Help Phone at 1-800-668-6868 or by texting CONNECT to 686868. The Hope for Wellness Helpline is available to Indigenous peoples across Canada 24/7 at 1-855-242-3310. If you or someone else is in immediate danger, call 9-1-1 or go to your nearest emergency department.