Can ChatGPT match physicians in the diagnosis, triage, management and prevention of infectious diseases? A scoping review

INTRODUCTION: The global burden of infectious diseases (ID), combined with a shortage of ID specialists, continues to challenge healthcare systems worldwide. Artificial intelligence (AI), particularly large language models such as ChatGPT, has emerged as a potential adjunct for diagnosis, triage, clinical decision-making, and patient communication. However, evidence describing its role in ID care and comparison with physician-led practice remains fragmented. To map and synthesize the current literature on the use of ChatGPT in infectious disease diagnosis, triage, and management, summarizing reported performance outcomes, clinical applications, and identified knowledge gaps in comparison with conventional physician-led care. METHOD: This scoping review followed PRISMA-ScR guidance. PubMed, Embase, and the Cochrane Library were searched inception in January 2026, with reference list screening. Peer-reviewed studies evaluating ChatGPT in ID-related contexts were included if they compared outputs with physicians, infectious disease specialists, residents, or established clinical standards. Outcomes of interest included diagnostic accuracy, triage, clinical management, antimicrobial prescribing, patient communication, screening, and education. Data were extracted using a Joanna Briggs Institute form and synthesized narratively. RESULTS: Fifteen studies were included, covering bloodstream infections, community-acquired pneumonia, tuberculosis, HIV, sexually transmitted diseases, histopathology, pharmacotherapy, emergency medicine triage, and screening. GPT-4 outperformed resident physicians in emergency diagnostic accuracy and generated generally appropriate management plans for pneumonia. In contrast, for bloodstream infections, ChatGPT achieved optimal management in only 1 of 44 cases, with potentially harmful recommendations reported in 16%. Patient-facing responses were often accurate and empathetic, particularly for HIV prevention, but frequently failed to meet recommended readability standards for sexually transmitted disease education. Antimicrobial and pharmacotherapy recommendations were commonly incomplete or discordant with expert practice. CONCLUSIONS: This scoping review provides an infectious disease-specific synthesis of evidence on ChatGPT across diagnosis, triage, management, antimicrobial stewardship, communication, screening, and education. The findings show that ChatGPT may support selected clinical and public health tasks but remains unreliable for complex infections and antimicrobial decision-making. Therefore, ChatGPT should be used cautiously as a supervised adjunct, not as a replacement for infectious disease clinicians. Future prospective studies and governance frameworks are needed to guide safe implementation

• Epidemiology and Determinants of Health
  • Determinants of Health
• Determinants of Health
  • Health services
  • Other
• Population(s)
  • General HIV+ population
  • General HIV- population
  • Other
• Prevention, Engagement and Care Cascade
  • Engagement and Care Cascade
  • Prevention
• Engagement and Care Cascade
  • Treatment
• Prevention
  • Education/media campaigns
• Co-infections
  • Other
• Health Systems
  • Governance arrangements
  • Delivery arrangements