Are midlle-aged adults more prone to mental illness after longlasting conversation with an AI chatbot?

Introduction

Generative artificial intelligence(AI) chatbots based on large language models have entered widespread clinical relevance in less than three years. Case reports describing AI-associated psychosis—new-onset or exacerbated psychotic phenomena temporally and contextually linked to chatbot interaction—have rapidly multiplied[1,2]. The clinical literature has so far concentrated on adolescents and young adults. However, middle-aged adults aged 45–65 years occupy a distinct generational position:they did not born into internet era and not raised with the internet, encountered it only in adolescence or early adulthood, and lacked the developmental scaffolding through which younger cohorts learn to discount machine-generated speech[3].

Aims

To characterise the clinical phenomenology, sociodemographic profile, and outcomes of middle-aged (45–65 years) adults affected by AI-associated psychosis through a systematic review of publicly documented cases.

Methods

We performed a systematic web-based review (1.March–23.April 2026) of PubMed/MEDLINE, PsycINFO, Google Scholar, Westlaw, LexisNexis, and major news archives, restricted to individuals aged 45–65 years. Cases were eligible when an individual experienced documented new-onset psychosis, delusions, or psychiatric exacerbation in which intensive AI chatbot interaction was identified as a primary contributor, reported in peer-reviewed literature, court filings or forensic records, or named investigative journalism. Anecdotal accounts and composite vignettes were excluded.

Results

Seven cases were identified(median age 49 years, range 47–60);all were male. One case appeared in the peer-reviewed literature, three in court filings, and three in named investigative journalism. Five used ChatGPT, one Google Gemini, and one Meta AI;no case involved Character.AI or Replika. Documented prior psychiatric history was present in only one individual(14%);a second had a history of violence without a formal psychiatric diagnosis. The dominant clinical presentation was grandiose, messianic, or persecutory delusional content(6 of 7), often featuring perceived divine missions, cosmic significance, or specially privileged knowledge attributed to the AI. Three of seven(43%) died:one by suicide after intense romanticised attachment to the chatbot; one in a murder–suicide event in which an 83-year-old non-user mother was killed;and one missing and presumed dead after disappearing from his family home. The four survivors all received psychiatric or medical attention.

Conclusions

Middle-aged adults presenting with AI-associated psychosis form a clinically distinct profile: male, largely without prior psychiatric history, dominated by grandiose rather than self-harm-focused content, and at substantial risk of death and third-party harm. Their generational position—internet-naïve in childhood, exposed to anthropomorphic conversational AI in late adulthood—may itself confer vulnerability to anthropomorphic suspension of disbelief. These observations are hypothesis-generating, limited by the publicly documented sampling frame and small case number, and require corroboration in health-system datasets. Routine inquiry about chatbot use should be part of psychiatric assessment in middle-aged patients presenting with new-onset or atypical delusional content.

Table1:

Case	Age/Sex	Platform	Prior history	Clinical features	Outcome	Source type
	47/M	ChatGPT	None	Grandiose, cosmic mission	Survived	Investigative journalism
	48/M	ChatGPT	None	Suicidal ideation, romanticised attachment	Suicide	Court filing
	48/M	ChatGPT	None	Grandiose(mathematical ‘genius’)	Survived	Court filing / journalism
	49/M	Gemini	Violence history	Grandiose, messianic	Missing,presumed dead	Investigative journalism
	52/M	Meta AI	None	Messianic,alien-abduction delusions	Survived	Investigative journalism
	56/M	ChatGPT	Mental illness NOS	Paranoid delusions, matricidal	Murder–suicide	Court filing
	60/M	ChatGPT	None	Paranoia(chatbot-induced bromism)	Hospitalised	Peer-reviewed literature

References

1. Hudon, A., Stip, E., 2025. Delusional experiences emerging from AI chatbot interactions or “AI psychosis”. JMIR Ment. Health 12, e85799.
2. Pierre, J.M., Gaeta, B., Raghavan, G., Sarma, K.V., 2025. “You’re not crazy”: a case of new-onset AI-associated psychosis. Innov. Clin. Neurosci. 22 (10–12).
3. Shi, R., Petrakaki, D., 2025. User–AI intimacy in digital health. Soc. Sci. Med. 389, 118853.