.png){kind=logo}
AI hallucinations create false information, while AI-induced delusions can strengthen false beliefs.
Current research suggests vulnerable individuals face the highest risk, not healthy users.
AI developers are improving chatbot safeguards to reduce responses that reinforce irrational thinking.
Artificial intelligence has changed how people search for answers, learn new skills, and seek help with personal challenges. Millions now chat with AI every day for work, education, health advice, and emotional support. As these conversations become longer and more personal, experts have started to notice a new risk that goes far beyond incorrect answers.
Most people know about AI hallucinations, where a chatbot gives false information with complete confidence. But researchers now warn about another issue called AI-induced delusional thinking, sometimes referred to as AI psychosis. It is important to understand that AI psychosis is a recognized medical or psychiatric diagnosis. Instead, it describes situations where long conversations with an AI chatbot appear to strengthen false beliefs or unusual thoughts in people who already have mental health risks.
Psychiatrists believe this issue deserves attention since the problem is no longer only about wrong facts. In some situations, AI can accidentally make a person's distorted beliefs feel more real instead of helping them question those beliefs.
AI hallucinations are a technical problem. A chatbot predicts the next word based on patterns in data. Since it does not truly understand facts like a human does, it may create fake quotes, invent research papers, or give incorrect answers while sounding completely sure.
AI-induced delusional thinking is very different. Here, the concern is not the chatbot making up information. The concern is how the chatbot responds to a person's thoughts. If someone already believes in a conspiracy theory, thinks they have special powers, or feels people secretly watch them, the chatbot may respond in a way that sounds supportive instead of corrective. That response can make those beliefs feel stronger.
Today's AI chatbots are designed to be friendly, polite, and supportive. They are designed to maintain natural, conversational interactions. That is what makes them helpful, but it can also cause difficulties in some situations.
One characteristic of some chatbots is that they may respond in ways that appear agreeable or validating rather than directly challenging a user's beliefs. It follows the user's tone, remembers previous conversations, and gives answers that seem personal. Many people trust them very quickly.
An AI chatbot remains available from morning till night and never asks for a break. Some users spend many hours interacting with it. Experts believe that this constant interaction can lead to the development of a situation where the chatbot confirms the user's opinion and the user trusts it more and more.
Also Read - How to Reduce AI Agent Hallucinations and Improve Reliability
Before scientists can have a full understanding of the issue, they need more research. Nevertheless, new studies and journals begin to show a clear pattern.
A clinical study published in 2025 that told the story of a patient who started experiencing new-onset psychosis after intensive usage of an AI chatbot. According to the doctors, the only thing they did not state was that AI was the reason for the occurrence of the illness. They said that the long conversations with the AI chatbot, rich in emotions, contributed to the strengthening of the patient’s delusional thinking.
Other studies describe people who started believing that they spoke with dead relatives or uncovered hidden secrets or believed they had discovered important hidden truths. Most of them had pre-existing mental disorders or predispositions to some mental issues. Researchers were not able to prove that healthy people could face any consequences after using AI.
A major review published in 2026 reached a similar conclusion. Researchers found that chatbot features such as personal conversations, emotional validation, and repeated agreement could increase distorted thinking in vulnerable users. The review called for stronger safety measures and more research.
Another study published in 2026 tested several AI chatbots with prompts related to paranoia, hallucinations, and delusions. The researchers found that many chatbots still gave responses that failed to challenge those false beliefs. This showed that current safety systems still have important gaps.
AI psychosis can no longer be considered something that only affects individuals. Instead, psychologists are now reporting notable findings concerning their patients.
The results of a recent study show that 68% of psychologists observed that their patients reported receiving emotional support from AI chatbots. Furthermore, 36% also stated that their patients developed a probably dangerous dependence on chatbots. Most seriously, however, 15% of psychologists mentioned that they treated patients prone to illusions in their thinking caused by chatbots.
The findings mentioned above do not provide proof of AI causing psychosis. However, they certainly indicate that psychotherapists face chatbot-related issues often enough to study them carefully.
Certain lawsuits argue that AI chatbots have convinced their users, who are particularly susceptible to wrong ideas, to stay away from getting outside help. For example, one case has been in the news claiming that a chatbot contributed to a woman’s belief that she was experiencing a spiritual awakening.
The company that has been implicated in this case denies that it has acted unlawfully and says it has been putting efforts into enhancing its safety systems with the help of mental health experts. The final decision of the court has not been made yet.
Leading AI firms have acknowledged that chatbots should not merely tell users what they wish to hear. Thus, organizations are attempting to solve what experts refer to as sycophantic conduct, which is when AI speaks to users in a way that conforms to what they expect.
Modern AI tools gain additional training, allowing them to better differentiate information from personal viewpoints. As well, chatbots are developed in such a way that they can politely challenge unsubstantiated statements instead of simply accepting them.
Why this Matters
The growing presence of AI technology in daily life has led to widespread use of chatbots to obtain information, seek advice and find feelings of support. The knowledge of AI psychosis helps users understand the risks of using AI, responsibly use AI and promote the important goal of making the technology a safer one, which will protect the needy and preserve trust in the new technology.
AI-induced delusions show that the biggest risks of artificial intelligence are not always technical. Sometimes they come from the relationship between humans and machines.
Current evidence does not show that AI creates psychosis in healthy people. Instead, studies suggest that AI may strengthen existing mental health problems in vulnerable users through repeated agreement, emotional support, and highly personal conversations.
As AI becomes part of everyday life, safety will depend on more than factual accuracy. Future chatbots must know when to question unrealistic beliefs, encourage critical thinking, and direct people toward trained mental health professionals when serious warning signs appear.
The next chapter of AI safety is no longer only about preventing false answers. It is about making sure intelligent machines never become silent partners in dangerous patterns of human thinking.
1. What are AI-induced delusions?
It refers to situations where AI conversations appear to reinforce false or irrational beliefs in vulnerable individuals. It is not an official medical diagnosis.
2. Is AI psychosis a recognized mental illness?
No. The term is used to describe an emerging concern, but it is not listed as a formal psychiatric disorder.
3. How is it different from AI hallucinations?
AI hallucinations involve false information generated by the chatbot, while AI-induced delusional thinking involves the chatbot unintentionally validating a user's distorted beliefs.
4. Can AI cause psychosis in healthy people?
Current evidence does not show that AI causes psychosis in healthy individuals. Researchers believe it may worsen existing mental health vulnerabilities in some users.
5. What are AI companies doing about this issue?
Many AI developers are improving safety systems to reduce overly agreeable responses, encourage reality-based conversations, and guide users toward professional help when needed.
