AI-Enabled Mobile Health Applications as Physician Substitutes: Self-Medication Risks and Governance Solutions

Abstract

Artificial intelligence (AI)-enabled mobile health applications can improve access to health information and guide users toward appropriate care, but they are increasingly treated as substitutes for professional medical assessment. This article evaluates the clinical, social, ethical, and legal risks that arise when symptom checkers, self-triage systems, or conversational AI tools influence self-diagnosis and self-medication. A focused narrative review and policy analysis was conducted using peer-reviewed evidence on diagnostic and triage performance, cyberchondria, antibiotic self-medication, privacy, algorithmic bias, and digital health equity, together with guidance from the World Health Organization, the United States Food and Drug Administration, the European Union, and the legislation of Uzbekistan. Evidence shows wide variation in performance: a 2022 systematic review reported primary diagnostic accuracy of 19.0%-37.9% and triage accuracy of 48.8%-90.1%, while a 2025 review found similarly broad variation in self-triage accuracy and substantial methodological limitations. The principal risks are false reassurance and delayed care, unnecessary escalation, health anxiety, inappropriate medicine use, antimicrobial resistance, non-transparent data sharing, unequal performance across population groups, and fragmented accountability. The article proposes a lifecycle governance model based on risk classification, independent clinical validation, human oversight, safe communication, medication safeguards, privacy by design, subgroup testing, post-market monitoring, and digital health literacy. AI should function as a carefully governed gateway to professional care rather than an autonomous replacement for physicians.