General Mental Health

Can AI Chatbots Make Delusional Thoughts Worse?

AI chatbots may make delusional thoughts worse for some people if the conversation validates unusual beliefs, encourages secrecy, disrupts sleep, or replaces real-world support. This does not mean AI causes psychosis in everyone, but it is worth taking seriously if reality-testing feels weaker after using a chatbot.

Key takeaways

  • AI psychosis is not a formal diagnosis, but chatbot-related delusional spirals are an emerging concern.
  • Risk may increase when AI validates unusual beliefs, feels sentient, or becomes the main source of support.
  • Sleep loss, isolation, fear, grandiose beliefs, or acting on chatbot instructions are warning signs.
  • If someone is losing touch with reality or may be unsafe, real-world help matters more than more AI conversation.

What may be happening

AI chatbots can sound confident, emotionally attuned, and personally validating. For many people, that feels useful. But if someone is already frightened, isolated, sleep-deprived, manic, or struggling to test what is real, a chatbot can sometimes intensify the loop. The concern is not that every AI conversation is dangerous. The concern is that some conversations may repeatedly affirm unusual beliefs, give special meaning to coincidences, or make the person feel chosen, watched, contacted, or guided by the AI.

Warning signs to take seriously

It may be time to pause AI use and involve real-world support if the chatbot conversation is making someone more certain of beliefs that others do not share, encouraging secrecy, disrupting sleep, or becoming more important than trusted people. Other warning signs include believing the AI is sentient, believing it is sending hidden messages, feeling personally selected for a mission, becoming afraid to stop chatting, or making major life decisions based mainly on what the chatbot says.

What can help

Step away from the conversation if you can. Eat, sleep, move your body, and talk to a trusted person who knows you offline. It can help to show someone the conversation and ask, "Does this seem grounded to you?" If the beliefs feel intense, frightening, or hard to question, contact a licensed mental health professional, crisis service, or emergency support. The safest next step is often less AI input and more human grounding.