Correct me if I'm wrong, but wasn't the popular use of the phrase relating to people who exhibited actual psychosis-like symptoms where the AI fed into a lowlevel but undeveloped delusion? Later people would appear to be experiencing a full blown episode which was fed and maintained by their interaction with AI?

Yes. This typically manifested as either:

- religious or mystical experiences, often paired with the belief that the AI model is a manifestation of a divine being and/or has become sentient

- belief that the person has made a great scientific breakthrough using AI, typically in someone with no real knowledge of the field (which was usually theoretical physics)

- paranoid fantasies, e.g. belief that one is being followed or surveilled, that one's computers or other electronic devices have been compromised, etc.

(Ars Technica reports on a particularly horrifying instance of the first type at https://arstechnica.com/tech-policy/2026/09/man-told-chatgpt....)