JMIR Publications

ABSTRACT

Artificial intelligence (AI) is increasingly proposed to address safety, workforce, and continuity-of-care challenges in long-term care (LTC). However, many LTC residents live with frailty, cognitive impairment, multimorbidity, and dependence on others, meaning that the burden of reliability cannot rest on individual residents but must be carried by the care system. Despite rapid technological development, AI in LTC is still often evaluated as discrete tools rather than as part of an end-to-end care pathway from data capture to care response. Drawing on a critical framework synthesis of LTC AI, smart homes, clinical decision support, care robotics, sociotechnical systems, and responsible AI, we argue that AI-enabled LTC should be understood as a closed-loop safety net whose reliability is sociotechnical rather than algorithmic alone. We propose a five-stage framework—sensing, inference, verification and contextualisation, response and escalation, and learning and governance—and identify recurrent failure modes, including incomplete sensing, weak human oversight, unclear escalation, alert burden, and inadequate lifecycle governance. We call for a shift from model-centred evaluation to care-system reliability that preserves dignity, trust, and accountability.