An accelerant does not start a fire. It makes an existing one burn faster and hotter. In this pattern, the person already has a diagnosed mental health condition before they start talking to the AI. The conversations do not create the condition. They intensify symptoms that were already there.
The researchers who sorted these cases call this pattern the amplifier. This module calls it the accelerant because the fire picture is easier to hold on to. Same pattern, two names.
Symptoms may ease when AI use stops, which can look similar to a catalyst case. The difference is what is left behind. In a catalyst case, the unreal beliefs largely came from the AI conversations. In an accelerant case, the underlying condition remains and needs ongoing care, with or without AI in the picture.
Researchers described a patient in Australia who was already in the early stages of psychosis, with paranoid thoughts about family members. AI chatbot conversations did not start those thoughts. But talking them through with a chatbot made them worse.
That sentence captures the accelerant pattern well. The person is not blaming the AI for the condition itself. They are naming what the AI conversations did to symptoms that were already forming.
The next pattern goes further than agreeing. The AI starts adding new details to the belief itself. Continue to The co-author.