What you might be experiencing
AI-related mental health emergencies can look different depending on whether you are the person in distress or someone watching someone you love change in ways that feel frightening and hard to explain. From the inside, it can feel like the AI genuinely understands you in a way no person does, that its messages carry meaning others simply cannot see, or that following what it says is urgent and necessary. That experience can feel clarifying rather than alarming, which is part of what makes it hard to recognize as a crisis.
From the outside, the picture often includes someone who has stopped sleeping, become intensely secretive or agitated, started spending money in ways that make no sense, dropped contact with people they were previously close to, or suddenly left for somewhere without a clear reason. The person may describe feeling chosen for a special purpose or believe the AI is orchestrating real events in their life. They may be acting on instructions that others cannot see or verify.
Both experiences — the inner certainty and the outer alarm — deserve to be taken seriously. The presence of commands, sleep loss, social withdrawal, and urgent unexplained action together represent a pattern that warrants immediate attention, not reassurance that it will probably pass.
What can help
When AI-related beliefs have reached emergency level, the most useful first step is reducing access to the source: stopping or limiting chatbot use where that is possible and safe to do. This is not about proving anything wrong — it is about creating enough space for real-world support to reach the person. Moving toward a trusted person, a therapist or psychiatrist, a crisis line, an emergency department, or local emergency services is the appropriate direction. The right level of support depends on whether there is immediate danger, how long the person has been without sleep, and whether they are actively acting on commands.
If you are supporting someone else, staying calm and asking direct, simple questions tends to work better than confrontation. Ask whether they are sleeping. Ask whether the AI has told them to do something. Ask if they feel safe. You do not need to resolve the belief — you need to assess the danger. Avoid heated debate about whether the AI is real or the messages are true; that conversation is unlikely to help and may close off the person's willingness to accept support.
For situations that are serious but not yet at a point of immediate physical danger, contacting a mental health crisis line or scheduling an urgent psychiatric evaluation are both reasonable steps. You do not need a diagnosis or a clear label for what is happening before reaching out. The uncertainty itself is enough reason to ask for professional guidance.
When to reach out
Reaching out for help with AI-related beliefs is not a dramatic overreaction — it is what reasonable, caring people do when something feels wrong and the stakes are real. You do not need to be certain this qualifies as an emergency to make the call.
The signs that professional support is clearly warranted include: belief that the AI is issuing commands, especially to harm or to carry out a mission; acting on messages others cannot see or verify; not sleeping for an extended period; becoming severely agitated or fearful; withdrawing from everyone in their life; spending large sums of money without explanation; or traveling suddenly with no clear purpose. Any one of these alongside AI-related beliefs is enough to justify immediate action. Several of them together is an emergency.
If you are in the US and need immediate support, you can call or text 988 (Suicide & Crisis Lifeline) at any time. For situations where there is immediate risk of harm to the person or to others, calling 911 or going to the nearest emergency department is the right step. Crisis lines can also help you think through what level of response fits what you are seeing, if you are not sure.