An Evidence-Based Framework for Patient-Facing Artificial Intelligence Integration in the Emergency Department

The emergency department is a high-need setting for patient-facing artificial intelligence (PF-AI) because it functions as both a health care safety net and, increasingly, a digital front door for unscheduled acute care. We define PF-AI as tools that interact directly with patients to collect information, deliver tailored communication, and support decision making under clinician governance. This article presents a constrained framework for where PF-AI may add value across the emergency care journey-including structured intake, patient education, multilingual communication, and postdischarge follow-up-and where strict boundaries are required given risks of undertriage, confabulated content, workflow burden, privacy harm, inequitable access, cost escalation, and medico-legal ambiguity. We map operational touchpoints from prearrival to follow-up, specify information-flow and accountability checkpoints, and propose a human oversight model with defined escalation roles. The framework emphasizes equity, interoperability, usability, privacy, safety verification, and auditability, recognizing that PF-AI should augment rather than replace clinician judgment or existing non-AI workflows.