What may be happening
AI tools can be useful, but they can also become part of a loop. You might notice that you feel calmer for a few minutes, then need to ask again. Or you may start choosing AI over sleep, friends, therapy, or your own judgment. For some people, AI conversations may also intensify unusual beliefs, grandiose ideas, fear, or the feeling that the chatbot has special knowledge about them. That deserves real-world attention.
Warning signs to take seriously
Practical warning signs include using AI for hours longer than intended, hiding use from people you trust, feeling worse after chats, depending on AI to make basic decisions, or using it to repeatedly check fears. Higher-risk signs include not sleeping, feeling commanded by AI, believing the AI is sentient or sending secret messages, hearing or seeing things others do not, feeling manic or invincible, or having thoughts of suicide or violence.
What can help
If the signs are mild, try a structured pause: set a time limit, remove late-night access, stop repeated reassurance prompts, and replace AI chats with one offline support step. If the signs are intense, frightening, or connected to safety, involve a trusted person, clinician, crisis line, or emergency service. Do not try to resolve reality-testing or crisis concerns by asking AI more questions.