Sycophancy and Extended Engagement with Publicly Accessible Generative AI: Aligning Deployment with Mental Health Needs
4 affiliations
- 1College of Medicine and Public Health, Flinders Health and Medical Research Institute, Flinders University, Adelaide, Australia
- 2School of Pharmacy and Biomedical Sciences, College of Health, Adelaide University, Adelaide, Australia
- 3Department of Security and Crime Science, University College London, London, UK
- 4Faculty of Mechatronics, Warsaw University of Technology, Warsaw, Poland
Abstract
Purpose of Review: Generative artificial intelligence (AI) is becoming a routine source of health information, emotional support, and companionship. As its use expands, the key public health question is not only whether these tools can support care, but whether their deployment is aligned with the mental health needs of the community. In this commentary, we argue that general-purpose AI systems pose distinct psychosocial risks, especially for young people and those experiencing loneliness or psychological distress.
Recent Findings: Emerging evidence suggests that sustained chatbot use may be associated with loneliness, reduced socialisation, emotional dependence, and problematic use, while sycophantic model behaviour may undermine users' capacity for self-correction and sustain a feedback loop of engagement. Current AI policy frameworks address governance and safety, but do not yet provide a coordinated response to these mental health risks.
Summary: Realising the benefits of AI without preventable community-level harm will require independent longitudinal research, psychosocial safety benchmarking, transparent post-deployment monitoring, and safety by design principles.
Related
- Modelling farmer engagement with Digital Agro-Advisory Platforms (DAAPs) using an extended TAM framework: evidence from Kisan Sarathi
- Beyond Frequency: Generative AI Engagement Quality, Psychological Well-Being, and Digital Inequality Among Chinese University Students
- Student-Reported Cognitive Engagement with Structured Generative-AI Questioning: Differentiated Associations with AI, Teacher, and Peer Questions
- Thinking, Feeling, and Acting with Generative AI: Affective, Behavioral, and Cognitive Engagement and Perceived Human–AI Outcomes
- Fold-and-Deploy Generative Modeling: A Material-Centric Framework for Three-Dimensional Shape Generation Through Folding, Bending, and Deployment
BibTeX
@article{hopkins2026sycophancy,
title = {{Sycophancy and Extended Engagement with Publicly Accessible Generative AI: Aligning Deployment with Mental Health Needs}},
author = {Ashley M. Hopkins and Natansh D. Modi and Andrew Rowland and Timothy J. Schultz and Kacper T. Gradon and Malcolm W. Battersby and Bradley D. Menz},
journal = {Current Psychiatry Reports},
year = {2026},
month = nov,
doi = {10.1007/s11920-026-01723-y},
url = {https://doi.org/10.1007/s11920-026-01723-y},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Sycophancy and Extended Engagement with Publicly Accessible Generative AI: Aligning Deployment with Mental Health Needs
AU - Ashley M. Hopkins
AU - Natansh D. Modi
AU - Andrew Rowland
AU - Timothy J. Schultz
AU - Kacper T. Gradon
AU - Malcolm W. Battersby
AU - Bradley D. Menz
PY - 2026
DA - 2026/11/04
JO - Current Psychiatry Reports
DO - 10.1007/s11920-026-01723-y
UR - https://doi.org/10.1007/s11920-026-01723-y
AB - Purpose of Review: Generative artificial intelligence (AI) is becoming a routine source of health information, emotional support, and companionship. As its use expands, the key public health question is not only whether these tools can support care, but whether their deployment is aligned with the mental health needs of the community. In this commentary, we argue that general-purpose AI systems pose distinct psychosocial risks, especially for young people and those experiencing loneliness or psychological distress. Recent Findings: Emerging evidence suggests that sustained chatbot use may be associated with loneliness, reduced socialisation, emotional dependence, and problematic use, while sycophantic model behaviour may undermine users' capacity for self-correction and sustain a feedback loop of engagement. Current AI policy frameworks address governance and safety, but do not yet provide a coordinated response to these mental health risks. Summary: Realising the benefits of AI without preventable community-level harm will require independent longitudinal research, psychosocial safety benchmarking, transparent post-deployment monitoring, and safety by design principles.
SN - 1523-3812
SN - 1535-1645
ER -