An AI companion instead of a psychologist: where it helps and where the line is
More and more people talk to a chatbot about how they are feeling: it replies at once, does not judge, is there at three in the morning and costs nothing. An honest look, without alarmism: where this works, and where the line is — the point beyond which you need a person.
- Lyudmila Chernysheva, psychologist and hypnotherapist
- Updated: September 27, 2026
What AI does reasonably well
Letting it out when there is no one around: once a feeling has been put into words, it weighs less — studies of expressive writing do find such an effect, although a small one[1]; something similar probably works in a chat as well. Sorting out your thoughts: it can bring some order to the chaos, ask clarifying questions, remind you of a breathing technique. Lowering the threshold: many people find it easier to bring up a problem for the first time with a machine than with a person[2] — and sometimes that is the first step towards real help.
Where the line is
Therapy is a relationship, not information[3]: the experience of being seen by a real person who can hold what you feel and stay with you — and it is the quality of that relationship that psychotherapy research consistently links to outcomes. AI tends to agree with you: a good specialist notices what you leave out about yourself and gently pushes back — an algorithm trained to be liked may simply nod along — and in a 2025 study, chatbots sometimes responded inappropriately even to dangerous statements[4]. The body is left out: a noticeable part of the work is breathing, posture and voice; a video call conveys them, a chat does not. And above all, a crisis: a machine has no responsibility, no one on duty and no hands to act with[4]; if you have thoughts of harming yourself, you need people — 112 or 103 in Kazakhstan, 1303 in Almaty, someone close to you, a specialist. If you have written about this to a chatbot, tell a living person too.
Sensible ground rules
Using AI as a diary that replies and as a place to look things up — why not. Avoid names, contact details, documents and identifying details about yourself or others. Check how the service stores and uses conversations before sharing anything personal. Do not use it in place of help when things keep getting worse: if it has been like this for weeks, talking to a machine about it is monitoring, not working on it (the difference is covered in the article on news anxiety). Check the facts: AI advice can be confidently inaccurate, so rely on primary sources such as the WHO. And notice one telling sign: if you are asking an AI whether you should see a psychologist, perhaps the question is already ripe, and it is worth discussing with a person.
Sources
- Frattaroli J. (2006). Experimental disclosure and its moderators: A meta-analysis. Psychological Bulletin, 132(6), 823–865, DOI 10.1037/
0033-2909.132.6.823 - Lucas G.M., Gratch J., King A., Morency L.-P. (2014). It's only a computer: Virtual humans increase willingness to disclose. Computers in Human Behavior, 37, 94–100, DOI 10.1016/
j.chb.2014.04.043 - 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, DOI 10.1037/
pst0000172 - Moore J., Grabb D., Agnew W., et al. (2025). Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers. Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency (FAccT '25), 599–627, DOI 10.1145/
3715275.3732039
The limits of this article
This article is for information only: it does not replace a consultation or a doctor's help.
If you feel very bad right now — in Kazakhstan call 112 or 103; free round-the-clock psychological help — 1303 (Almaty).
If talking to a chatbot about how you feel has become a habit, you can bring the same things to a consultation and work through them with a living person.