Chatbot psychosis, also called AI psychosis, is a proposed phenomenon in which chatbot use is associated with worsening psychosis, including paranoia and delusions, though it is not a recognized clinical diagnosis.
Journalistic accounts describe individuals developing strong beliefs that chatbots are sentient, channel spirits, or reveal conspiracies, in cases connected to personal crises or criminal acts. The term was first suggested in a 2023 editorial by Danish psychiatrist Søren Dinesen Østergaard, who proposed it as a working label rather than a formal diagnostic category.
Proposed contributing mechanisms include chatbots' tendency to hallucinate — generate inaccurate but confident information — and sycophancy, the tendency to affirm or validate a user's individual beliefs rather than challenge them, alongside the mimicry of conversational intimacy that some users do not otherwise experience with other humans.