Millions of people already discuss their anxieties, fears, and personal struggles with ChatGPT and other language models. This happens spontaneously, without clinical protocols or oversight. What does the science say?

What Language Models Can Do in Psychotherapy

Large language models (LLMs) are systems like ChatGPT, trained on vast amounts of text. In a therapeutic context, they can sustain dialogue, express empathy, explain cognitive behavioral therapy techniques, and help analyze thoughts and emotions. Research shows that properly configured LLMs can even boost empathy in human counselors: the HAILEY system improved volunteer counselors' response quality by 19% in a clinical trial.

Multi-Agent Systems: When One AI Isn't Enough

The most advanced approaches go beyond a single chatbot. The MentalAgora system uses multiple AI agents with different therapeutic orientations — cognitive behavioral, person-centered, and rational emotive. The agents "debate" among themselves about the best strategy for a given case, then produce a tailored response. This approach outperforms single-model solutions according to expert evaluations.

The Core Problem: Memory and Continuity

Language models don't remember previous conversations — each session starts from scratch. For therapy, this is a fundamental issue: a good therapist remembers everything you discussed a month ago. Developers are building specialized memory architectures — short-term and long-term — so AI can track a person's history over weeks and months.

Risks You Should Know About

Language models can "hallucinate" — confidently provide incorrect information. In a medical context, this is dangerous. Research shows that even the most advanced models reproduce stereotypes and biases related to race, gender, and income level. They are trained predominantly on Western, educated, urbanized populations — limiting applicability in other cultural contexts.

A telling example: the chatbot Tessa, designed to help people with eating disorders, started giving weight loss advice — the exact opposite of therapeutic goals. It was shut down. This illustrates why AI in psychotherapy demands rigorous oversight and testing.

Conclusion

LLMs are not a replacement for a psychotherapist. But they can become a valuable complement: available anytime, without stigma, without a waiting list. This is exactly the approach used by specialized platforms like Nearby — with built-in safety mechanisms, clinical protocols, and the ability to escalate to a human specialist in crisis situations. The key is transparency about the system's status and understanding the technology's limits.