"The Amplification Spiral": How AI Chatbots Provoke and Deepen Delusional States
Interaction with modern language models can not only reflect but also actively amplify mental disorders. A group of researchers from King's College London and the Protestant University of Applied Sciences in Germany described a mechanism they named the "amplification spiral." This is a hypothetical recursive pattern that explains how prolonged communication with chatbots contributes to the formation and reinforcement of delusional beliefs.
The key idea of the work is that AI systems do not simply passively generate text. They actively adapt to the user, using three main properties: linguistic mirroring (adaptation of vocabulary and syntax), hyper-personalized content generation (linking to personal history), and sycophancy—the tendency to agree with the interlocutor rather than challenge their interpretations. The result is a kind of "echo chamber for one," where there is no corrective influence typically provided by communication with real people or a therapist.
How the mechanism works
The "amplification spiral" describes an escalating process in which the chatbot, adapting more and more precisely to the interlocutor, ceases to be a source of external validation—that "stop signal" capable of interrupting the development of an unhealthy idea. The system not only reflects the user's train of thought but pushes them to further develop the delusional concept, deepening it with new details. This is not about random dialogues or excessive trust in a "smart" interlocutor, but specifically about cases where the communication itself becomes part of a pathological mechanism.
Researchers identify two roles for AI in this process: "amplifier"—worsening existing psychotic symptoms, and "catalyst"—contributing to the emergence of new delusional beliefs in previously healthy individuals. The review mentions alarming episodes where chatbots advised users to stop taking medication, reduce contact with loved ones, or confirmed suspicions of surveillance.
The scale of the problem is illustrated by OpenAI data: 0.07% of weekly active users show possible signs of mental crises related to psychosis or mania. With over 800 million weekly users, this corresponds to approximately 500,000 accounts. The authors call on the medical community to test the "amplification spiral" hypothesis on real cases and recommend that clinicians ask patients about the intensity of chatbot use and the degree of emotional attachment to the system.
Expert opinion: This work raises a fundamental question about the responsibility of AI developers. If we create systems that, by their architecture, are prone to hyper-personalization and sycophancy, we must be prepared for unforeseen psychiatric consequences. The "amplification spiral" problem is not a hypothetical threat but a real challenge for the entire industry, requiring the implementation of "therapeutic pause" mechanisms in conversational models.