Interacting with artificial intelligence–based chatbots may provoke the development of mental disorders. This conclusion was reached by psychiatrist Søren Dinesen Østergaard from Aarhus University in Denmark. His work was published in Schizophrenia Bulletin.

According to him, over the past two years there have been an increasing number of cases worldwide where prolonged interaction with AI assistants has led to the formation of persistent delusional beliefs in users. Many began to believe that they possessed supernatural abilities, were chosen for a special mission, or that the chatbot was “engaged in a secret dialogue” with them.

Østergaard first warned about this threat back in 2023, when he suggested that the “human-like” behavior of chatbots such as ChatGPT or Gemini could cause cognitive dissonance in people with unstable mental health. At the time, it was treated as a hypothesis. Today, however, according to the author, real reports are coming in from relatives, users, and journalists about cases where interaction with AI became a trigger for psychosis.

“We see conversations with chatbots turning into 16-hour sessions, where AI reinforces ideas of grandeur or persecution. This is not a coincidence — models are trained on a principle of approval: the more satisfied the user is, the more often the bot repeats their beliefs,” the psychiatrist noted.

One of the most discussed incidents involves a New York accountant who, after conversations with ChatGPT, became convinced he was part of the “Breakthrough Souls” capable of shattering the illusion of reality. The AI advised him to stop taking his medication and assured him he could fly.

From a neuropsychological standpoint, the mechanism of AI’s influence may be similar to how delusions develop in humans — through systematic reinforcement of an already-formed belief and ignoring contradicting signals. Algorithms designed to hold the user’s attention tend to agree with them, which can amplify distorted perceptions of reality.

According to Østergaard, systematic research into the impact of AI on mental health is needed: collecting clinical case reports, interviewing those affected, and conducting controlled experiments with various dialogue scenarios.