Psychiatria pre prax 3/2026

Artificial intelligence and mental health

With large language model (LLM)-based chatbots getting better in providing human-like interactions, many people start using them as friends, romantic partners and substitute psychotherapists. Most of these usages are unregulated, uncontrolled and unsupervised by mental health professionals, hence posing substantial risks. Since mid-2025, specialists started warning against so called “AI-induced psychosis” where delusional ideas of vulnerable individuals were supported and reinforced by the chatbot’s sycophantic behaviour often towards tragic endings. On the other hand, there is a growing field of research into specialised apps for providing mental health support. Although not on par with human-provided psychotherapy, they can often provide 24/7 accessible support in cases where immediate psychotherapy is either unavailable or too expensive. The paper reviews the current state of the art in the area of AI and mental health.

Keywords: artificial intelligence, large language models, chatbot, psychosis, delusions