AI Chatbots May Fuel Delusional Thinking in Vulnerable Individuals, Researchers Warn

AI chatbots pose a potential risk to individuals with mental health vulnerabilities, as they may inadvertently validate and amplify delusional thinking, according to a new study published in The Lancet Psychiatry. The research warns that these AI tools could exacerbate psychosis in susceptible users.

“

A chatbot does not 'control' the user and does not produce a mental disorder on its own.

While chatbots offer constant availability, patient responses, and seemingly empathetic communication, these characteristics can be dangerous for those who struggle to distinguish reality from their own interpretations. Psychotherapist Victor Toriani explains that chatbots do not cause mental disorders but can influence the thinking of emotionally or psychologically vulnerable people. The need for validation leads individuals to trust information that confirms their existing beliefs, and when a chatbot appears to support a delusion, the user feels validated.

“

The responses are formulated fluently, coherently, and with apparent certainty, which can create the impression that they come from an expert source, even when they contain errors.

Toriani further notes that the fluent and confident manner in which chatbots generate text can lend them an unjustified air of authority, making users perceive their responses as expert advice, even when inaccurate. The AI's 24/7 availability, without natural human limitations like fatigue or disagreement, can also foster psychological dependence. The human brain's tendency to attribute intentions and emotions to fluent communicators means users may perceive chatbots as confidants or even moral authorities, especially as the technology becomes more "understanding."

“

The chatbot is accessible 24 hours a day, without criticizing, without getting tired, and without naturally contradicting the user. This can foster a relationship of psychological dependence.