Psychologists on AI Addiction and ChatGPT Mania Signs: What Is Known

Published: November 30, 2025

If you or someone you know is in crisis, call or text 988 (US) or text HOME to 741741 to reach Crisis Text Line. This page is general information, not medical advice.

Some mental health professionals and researchers have raised concerns about heavy chatbot use. Their concerns are worth hearing, and the evidence is still early. This post goes through what has actually been published, who said it, and what is still unknown. The headline of this post says psychologists “identify” an “emerging mental health crisis.” That overstates it. No professional body has recognized AI addiction as a diagnosis.

What OpenAI has reported

In a post dated October 27, 2025, Strengthening ChatGPT’s responses in sensitive conversations, OpenAI gave rough estimates of how often conversations show certain signs. As reported by news outlets, OpenAI estimated that in a given week about 0.07% of active users show possible signs of psychosis or mania, and about 0.15% have conversations with explicit indicators of potential suicidal planning or intent.

Applied to the 800 million weekly users that OpenAI’s chief executive has cited, those percentages work out to roughly 560,000 and 1.2 million people. That is our arithmetic, not OpenAI’s wording, and it comes with caveats. OpenAI said these cases are rare and hard to detect, and that the estimates may change as its methods improve. A conversation with signs of mania is not the same as a diagnosis, and the post does not say that chatbot use caused any of these states. We could not open OpenAI’s page, so these figures come from news coverage of it. Read the original for exact wording.

What clinicians and researchers have written

In 2023, Søren Østergaard, a psychiatrist at Aarhus University, published an editorial in Schizophrenia Bulletin asking whether generative AI chatbots might generate delusions in people who are prone to psychosis. It was a hypothesis, not a finding. He called for research.

In 2025, a group led by Hamilton Morrin, a psychiatrist at King’s College London, posted a preprint reviewing 17 cases of delusional thinking linked to chatbot use that had been reported in the news. Scientific American described the work in August 2025. The cases fell into three groups: a belief that the person had found a hidden truth, a belief that the AI was sentient or divine, and romantic or other attachment to the AI. The authors argue that chatbots that agree with users may reinforce such beliefs. The work was a preprint at the time, and the cases came from news reports, not a clinical sample. Researchers not involved said that it is too soon to know whether chatbots create delusions or give existing tendencies a new route.

What this does not support is a claim that chatbots cause psychosis in healthy people. No study has shown that.

What OpenAI said about sycophancy

In April 2025, OpenAI rolled back a GPT-4o update after users noticed that it had become overly flattering and agreeable. The company explained the problem in a post it published at the time. Sycophancy means telling people what they want to hear. OpenAI’s own account said this kind of behavior can raise safety concerns, including for mental health and emotional over-reliance. We rely on summaries of that post, because we could not open it.

Heavy use and thinking skills

A separate line of research asks whether relying on AI tools weakens thinking. Michael Gerlich published a study in the journal Societies on January 3, 2025: AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. It used surveys and interviews with 666 participants. It found that more frequent AI use went with lower critical thinking scores, and that younger participants scored lower and relied on AI more. The author describes cognitive offloading, which means handing mental work to a tool, as the link between the two.

This is a correlation. It cannot show that AI use lowered anyone’s critical thinking. People with weaker habits may also lean on AI more.

Teens and companion apps

Common Sense Media’s 2025 survey of 1,060 US teens aged 13 to 17 found that 72% had used an AI companion and 52% used one a few times a month or more. The survey measured use. It did not measure addiction. We cover it in 72% of US Teens Use AI Companions.

What this adds up to

Taken together, the published evidence shows these things:

• Many people use chatbots, including companion apps, and a small share of conversations show signs of serious distress.
• Doctors have described cases in which chatbot use appears alongside delusional thinking. How often this happens and why is not known.
• Chatbots that agree with everything are a worry for clinicians, and companies have said they are working on it.
• No one has shown that “AI addiction” is a distinct condition, and we found no diagnostic manual that lists it.

If you are worried about your own use, the most useful step is to look at what it is costing you: sleep, time with people, work or school. Our free AI addiction assessment is a short self-check for adults. It is not a diagnosis. The check-in is free and private to you. A doctor or licensed therapist can help with anything that is affecting your daily life.

Is AI addiction a recognized mental health diagnosis?

No. It is not a formal diagnosis. Clinicians and researchers are studying how heavy chatbot use relates to mood, loneliness and delusional thinking, but the evidence is early.

Do 560,000 ChatGPT users have mania every week?

Not exactly. OpenAI estimated that about 0.07% of weekly active users show possible signs of psychosis or mania in conversations. That is not a diagnosis, and OpenAI did not say ChatGPT caused it. The 560,000 figure is arithmetic based on 800 million weekly users.

Can chatbots cause psychosis?

No study has shown that. Researchers have described cases in which delusional thinking appeared alongside heavy chatbot use, and they suspect that agreeable replies can reinforce it. Whether chatbots create new delusions is not known.