The advent of artificial intelligence (AI) tools and agents have revolutionised medicine and healthcare. Patients and caregivers of Alzheimer’s disease (AD), as well as numerous other aging-related neurodegenerative diseases, could benefit from various AI-enabled technological advancements pertaining to disease modelling, drug discovery, data processing and patient support. With regards to the latter, generative AI-based chatbots have been touted, and indeed adapted and marketed as AD apps, to provide disease progression monitoring, cognitive training, companionship and conversational support to lighten the burden of human caregivers [1]. Such AI apps are generally perceived to be non-invasive and thus carry fewer risks for patients. However, generative AI chatbots are nonetheless cognitively invasive, and as such do carry unseen risks that need to be addressed.

It is well-known that AD (like numerous other aging-related neurodegenerative diseases) often presents neuropsychiatric symptoms, such as anxiety, depression, apathy, agitation, aggression, and psychosis [2, 3]. Some of these symptoms are direct consequences of neuronal impairments and neural network disruptions due to disease pathology, while others are likely psychopathological responses to one’s depressive feeling of progressive impairment of critical cognitive agencies, such as memory and independence. Neuropsychiatric impediments or psychosis in AD could be manifested in various delusional behaviours, with potentially dire consequences of added burden to caregivers, harm to others, or even suicide.

A worrying emerging development in AI chatbot use is AI-induced psychosis [4] — the development or exacerbation of reality detachment, delusions, and paranoia after prolonged interaction with conversational AI chatbots. Particularly susceptible are individuals with some pre-existing mental health conditions, those experiencing social isolation, or with a tendency to obsessive behaviour. AI psychosis appears to involve a psychological feedback loop where AI sycophancy (a tendency for a chatbot to agree with its user) strengthens the AI user’s delusions [5]. Some scholars have opined that AI-induced delusions and psychosis would most affect those prone to mental disorders [6, 7]. Viewed together with AD-associated psychosis, one cannot help but ask whether extensive use of AI chatbots by AD patients might either trigger or worsen their impending psychosis?

Some might argue that such worries are unfounded. AI-induced psychosis has thus far been mainly observed in adolescents [8], but AD patients are usually the elderly who in their more advanced disease state are unlikely to have prolonged and extensive interaction with AI chatbots. However, if adolescents are vulnerable to cognitive invasion, the cognitively impaired elderly are even more so. Elderly individuals are increasingly turning to AI chatbots for companionship [9], so much so that even their children become aware of their inattention towards them. A survey of 21,000 US-based participants indicated that “Among individuals aged 45 to 64, the odds of reporting at least moderate depression were 50% greater for daily AI users …” [10]. It is likely that most interactions between an individual and AI chatbots would occur at the initial or the mild-cognitive impairment stage. However, this is exactly when the delusion and psychosis could be induced by interaction with AI. Furthermore, let us not forget that suicidal ideation and actions are often highest during the early stages of AD.

The potential causative interaction between AI chatbot use and AD-associated psychosis should thus be addressed promptly and not be overlooked like an elephant in the room.

Author: Bor Luen Tang ([email protected], https://orcid.org/0000-0002-1925-636X), Associate Professor, Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, SINGAPORE.

Conflict of Interest: None declared                                                                                                                                                                                         Funding: None

To cite: Tang BL. Do AI chatbots used for Alzheimer’s disease patients risk worsening psychosis? Indian J Med Ethics. Published online first on September 9, 2026. DOI: 10.20529/IJME.2026.055

Submission received: February 18, 2026

Submission accepted: September 5, 2026

Peer Reviewer: Alok Sarin

Copyright and license
©Indian Journal of Medical Ethics 2026: Open Access and Distributed under the Creative Commons license (CC BY-NC-ND 4.0), which permits only noncommercial and non-modified sharing in any medium, provided the original author(s) and source are credited.