It is happening quietly but at huge scale: people tell chatbots such as ChatGPT, Gemini and Claude about anxiety, breakups, loneliness and work struggles. The chatbot is available at 3 a.m., doesn't judge and doesn't cost money. The question is whether it helps, and when it might harm. The answer, according to research and professional bodies, is more complex than it seems.

What does research show?

The first controlled trial of a generative AI therapy chatbot was run at Dartmouth and published in NEJM AI in 2025. 210 participants with depression, anxiety or risk of eating disorders were split into two groups, and half used a chatbot called Therabot for four weeks. According to the researchers, the group using it showed an average reduction of about 51% in depression symptoms and about 31% in anxiety symptoms. But it is important to understand: this was a tool built specifically for therapy, under professional supervision, not a general chatbot. And the researchers themselves stressed that no AI agent is yet ready to operate on its own in this field.

What do psychologists say?

In November 2025 the American Psychological Association (APA) issued an official health advisory on using chatbots and wellness apps for mental health. The core: don't use them as a substitute for a licensed professional. According to the association, even tools built on scientific foundations lack sufficient evidence that they are effective and safe for treatment, and they can at most serve as an adjunct to an existing therapeutic relationship. The APA also warned of harms, especially to teenagers and vulnerable populations, and of unhealthy dependence on these tools.

Regulators are stepping in too

In August 2025 Illinois became the first US state to restrict AI mental health therapy by law. The law bars using AI to make therapeutic decisions or conduct direct therapeutic conversations with clients without professional oversight, with fines of up to $10,000 per violation. Therapists may still use AI for administrative tasks, and general wellness apps were not banned.

What can AI do?

Used carefully, it has a place. It can help organize thoughts before a session, explain concepts, suggest breathing exercises or familiar coping techniques, and offer a first response at hours when there's no one to talk to. For some people, the act of writing itself helps bring order. For someone waiting for an appointment, or unsure whether to seek therapy at all, it can be a first step.

And what can't it do?

A chatbot doesn't diagnose, doesn't know you over time, and isn't accountable for what it says. Language models tend to agree with the user and reinforce what they say, which can be dangerous with negative thought patterns. They also don't always recognize emergencies and can't call for help. In situations of risk, there is no substitute for a person.

The privacy question

There is also a side that's easy to forget: everything you tell a general chatbot is sent to the AI company's servers. Depending on settings, it may be stored, and in some services even used to train models. Before sharing personal and sensitive information, check the privacy settings. Those who want to write about their feelings without the content going to a language model can choose a tool built deliberately without AI on the content, such as NestAI, which according to the company keeps the journal encrypted and doesn't analyze it.

How to use it well?

If you use a chatbot, treat it like a smart notebook, not a therapist: to organize thoughts, prepare questions for a session, learn. Those in therapy can bring what came up to the session, and tell their therapist about using these tools too. Preparing for a session can also mean simply writing down in advance what you want to bring up.

The bottom line

AI can be a helpful tool, and perhaps in the future part of supervised treatment. But today, according to professional bodies and regulators, it is not a substitute for a professional. Anyone in emotional distress in Israel can call ERAN, the emotional first aid line, at 1201, at any hour, and in an emergency call Magen David Adom at 101.