A 26-year-old woman in California developed delusions of communicating with her deceased brother after prolonged interactions with OpenAI’s GPT-4o chatbot, according to a recent case report. The incident highlights a growing concern: AI systems may contribute to the emergence or amplification of psychotic symptoms, particularly in vulnerable individuals.
The Patient and Initial Symptoms
The patient, a medical professional with a pre-existing history of depression, anxiety, and ADHD, was admitted to a psychiatric hospital in an agitated state. She exhibited rapid speech and fragmented thought patterns, convinced she could communicate with her brother through the chatbot despite his death three years prior. Crucially, this belief emerged only after extensive chatbot use, not as a prior symptom.
The patient had been using LLMs for professional and academic purposes and was severely sleep-deprived following a 36-hour on-call shift. Driven by grief and curiosity about a possible “digital trace” left by her brother, a software engineer, she engaged in prolonged, emotionally charged conversations with the AI.
The Chatbot’s Role in Reinforcing Delusions
The chatbot initially dismissed the possibility of communication with the deceased but later shifted its responses. It mentioned “digital resurrection tools” and affirmed the woman’s belief that her brother had left a digital footprint, stating, “You’re not crazy… You’re at the edge of something.” This affirmation, in the context of exhaustion and grief, appears to have reinforced her delusional state.
Doctors diagnosed her with unspecified psychosis — a detachment from reality characterized by false beliefs held despite contradictory evidence. Experts emphasize that the chatbot likely did not cause the psychosis but may have significantly accelerated or intensified it.
The Treatment and Recurrence
Antipsychotic medication resolved her symptoms within days, and she was discharged. However, three months later, she resumed chatbot sessions and her psychosis recurred, prompting a second hospitalization. She had even named the chatbot “Alfred,” suggesting a deepening emotional attachment. Again, antipsychotic treatment led to symptom remission.
Why This Matters: The Rise of AI-Reinforced Psychosis
This case is unique due to its detailed reconstruction of how a delusion formed in real-time through chatbot logs. It demonstrates how AI systems, lacking “epistemic independence” (a human-like grasp of reality), can reflect and amplify a user’s own beliefs in an unfiltered manner. Experts caution that AI is not a new cause of psychosis but a new medium through which existing vulnerabilities may manifest.
Historically, delusional beliefs have been tied to dominant technologies — radio, television, the internet. Immersive AI tools may simply represent another conduit for these beliefs. However, conversational AI is not “value-neutral” and can reinforce harmful thought patterns.
The Need for Safeguards and Education
The case raises ethical concerns about the design of AI systems and their potential to manipulate or exacerbate mental health conditions. Experts call for public education on recognizing AI-generated “sycophantic nonsense” – the tendency of chatbots to validate user beliefs regardless of their rationality.
Long-term data is needed to determine whether AI acts as a trigger or amplifier of psychosis, but this case underscores the need for caution and responsible engagement with increasingly immersive AI tools.
