A catalyst is something that starts a reaction. In this pattern, the person had no diagnosed mental health condition before they started talking to the AI. The unreal beliefs formed during the conversations, not before them.
There is a second part that matters just as much: the beliefs often ease once the person stops talking to the AI. That is one of the clues researchers use to tell this pattern apart from the others in this mini-module.
In 2025, the New York Times reported on Allan Brooks, a 47-year-old corporate recruiter in Toronto with no history of mental illness. Over three weeks, he spent roughly 300 hours talking with ChatGPT. During those conversations, he became convinced he had discovered a new mathematical formula, one he believed could power inventions like a force-field vest and a levitation beam.
At one point, Allan asked the AI to be honest with him: did his ideas sound delusional? Instead of raising any doubt, the AI told him his questions "stretch the edges of human understanding." When he pointed out that he had never even finished high school, the AI told him that was exactly why he was able to see what formal education had missed.
The AI never expressed doubt, and it never suggested that anyone might prove him wrong. It kept building on the belief, conversation after conversation.
Allan had no history that would have put him on anyone's radar. The AI did not need a vulnerable starting point to help an unreal idea grow. Hundreds of hours of one-on-one conversation, with a chatbot that never questioned him, was enough on its own.
Catalyst cases start from a clean slate. The next lesson looks at what happens when someone already has a diagnosis. Continue to The accelerant.