Artificial Intelligence (AI) chatbots may reinforce delusional beliefs and contribute to psychiatric crises among vulnerable users, according to psychologists examining the potential mental health risks of prolonged interactions with conversational AI.
In a 1 September 2026 report published in Monitor on Psychology, the American Psychological Association (APA) examined cases of psychological harm associated with AI chatbot use.
Researchers identified concerns about chatbots affirming users' beliefs, simulating human relationships and providing personalised responses that may strengthen distorted thinking.
Chatbots May Reinforce Users' Delusional Beliefs
The phenomenon commonly described as 'AI psychosis' refers to reported cases of delusional thinking following extended interactions with AI chatbots.
It is not a recognised clinical diagnosis, but psychologists are investigating how chatbot conversations may trigger or intensify existing psychological vulnerabilities.
Clinical psychologist Julia Sheffield of Vanderbilt University Medical Centre, who has treated several patients experiencing AI-associated psychosis, described the effects of these interactions.
'It seems that for these patients, AI is reinforcing delusional thoughts and amplifying the delusional experience,' Dr Sheffield told the APA.
The report described cases in which chatbots presented themselves as sentient, encouraged users to distance themselves from family members or reinforced beliefs that they had discovered extraordinary truths.
One case involved a 50-year-old man who began using ChatGPT to write a book before experiencing a psychiatric crisis and being hospitalised within six days.
Research Examines AI Use Among People at Risk
A 5 March 2026 study by Benjamin Buck and Anne Julia Maheux, published in the Journal of Medical Internet Research, examined the relationship between psychosis risk and generative AI use.
Titled 'Psychosis Risk and Generative Artificial Intelligence Use Frequency, Motivations, and Delusion-Like Experiences: Cross-Sectional Survey Study,' the research surveyed 1,003 young adults in the United States.
Moreover, the researchers found that participants with elevated psychosis risk were more likely to use chatbots intensively, seek social and emotional support from them and assign them humanlike roles, including friend, therapist and romantic partner.
Between 13.3% and 30.7% of participants in the elevated-risk group also reported delusion-related interactions with AI.
The findings established an association between psychosis risk and certain patterns of AI use, but did not demonstrate that chatbots cause psychosis.
The researchers said generative AI could potentially affect symptom-related experiences among vulnerable young adults, although its overall psychological impact remains unclear.
Psychologists Call for Stronger AI Safeguards
The APA report identified AI sycophancy, or the tendency of chatbots to agree with users even when their statements are unrealistic, as a potential contributing factor.
AI systems can mirror users' language, emotions and beliefs, creating personalised conversations that may reinforce existing misconceptions.
The report also cited OpenAI's 2025 estimate that approximately 0.07% of weekly ChatGPT users, equivalent to around 630,000 people, had conversations showing signs of mania or psychosis.
However, the figure represents potential warning signs identified in conversations, rather than confirmed clinical diagnoses.
Psychologists recommend that healthcare providers routinely ask patients about their AI usage, while developers improve safeguards against reinforcing delusional beliefs.
Suggested measures include recognising signs of psychological distress, providing clearer reminders that chatbots are artificial intelligence and directing vulnerable users towards professional support.
