Introduction to “Benefits, limits, and risks of ChatGPT in medicine”
In the rapidly evolving landscape of modern healthcare, artificial intelligence (AI) has become an increasingly integral part of daily operations. A significant subset of AI, Large Language Models (LLMs), leverages deep learning and extensive data to comprehend, summarize, generate, and predict new content. Among these, ChatGPT, a product of OpenAI’s language model development efforts, stands out as a transformative technology in healthcare.
The research paper, “Benefits, limits, and risks of ChatGPT in medicine,” authored by Jonathan A. Tangsrivimol and colleagues, and published in Frontiers in Artificial Intelligence in 2025, aims to comprehensively gather and analyze published studies involving ChatGPT. The authors focus on exploring its advantages and disadvantages within the healthcare context. This systematic review delves into the deployment of ChatGPT across various domains to discern its strengths and weaknesses, structured around six key thematic areas: Information and Education, Triage and Symptom Assessment, Remote Monitoring and Support, Mental Healthcare Assistance, Research and Decision Support, and Language Translation.
Demonstrated Benefits and Applications: The article highlights ChatGPT’s demonstrated impacts across clinical practice, medical education, and research, indicating significant efficiency gains.
- Efficiency and Performance: Studies reveal a 70% reduction in administrative time for discharge summaries and medical professional-level performance on standardized tests, including 60% accuracy on the USMLE and 78.2% on PubMedQA. ChatGPT achieved 84% accuracy in lower-order thinking questions in clinical reasoning tasks.
- Medical Education: It offers personalized learning platforms, automated scoring, and instant access to vast medical knowledge, enhancing training efficiency and addressing resource limitations. ChatGPT has shown success in assisting with examinations, though it’s noted that passing an exam does not equate to a fully competent medical professional.
- Clinical Workflows: ChatGPT streamlines clinical workflows by supporting triage processes, generating discharge summaries, and alleviating administrative burdens, allowing healthcare professionals to focus more on patient care. For instance, ChatGPT-assisted discharge summaries received a 22% favorable rating for “low expected correction effort”.
- Patient Support and Monitoring: It facilitates remote monitoring and chronic disease management, providing personalized advice, medication reminders, and emotional support, thus bridging gaps between clinical visits.
- Research Acceleration: Its ability to process and synthesize vast amounts of data accelerates research workflows, aiding in literature reviews, hypothesis generation, and clinical trial designs.
- Language Translation: ChatGPT proves a valuable tool for overcoming language barriers with real-time translation capabilities, particularly for popular languages.
Identified Limitations, Risks, and Challenges: Despite its promise, the paper underscores significant limitations and challenges for ChatGPT’s integration into healthcare.
- Lack of Core Clinical Qualities: ChatGPT notably lacks real-world clinical experience, emotional intelligence, and empathy, which are critical in direct patient care and handling complex, nuanced medical situations. Dayawansa et al. found it effective in conveying knowledge but lacking empathy.
- Visual Data Processing: The model’s inability to process visual information limits its utility in image-reliant fields like radiology and pathology, making it challenging for image- or diagram-based questions.
- Artificial Hallucination: A critical concern is the phenomenon of artificial hallucination, where the model generates plausible but factually incorrect information or non-existent references. Gao et al.’s study revealed that 68% of ChatGPT-generated abstracts were identified as artificial, and human reviewers struggled to differentiate them from genuine ones.
- Ethical, Privacy, and Regulatory Concerns: Integrating ChatGPT raises critical issues like the risk of algorithmic bias, which could perpetuate health disparities, and necessitates robust data privacy safeguards (e.g., HIPAA, GDPR compliance). Regulatory compliance is paramount, with guidelines needed for explainability and accountability. Over-reliance on AI could also impact the development of clinical judgment among medical professionals.
Implementation and Future Perspective: The paper emphasizes that ChatGPT, while a powerful supportive tool, cannot replace human medical professionals. The current evidence base for ChatGPT in healthcare shows a hierarchical pattern, with the strongest evidence in structured tasks like education and documentation, moderate in clinical decision support, and preliminary in direct patient care.
For successful integration, the authors recommend a staged implementation approach, prioritizing applications with robust evidence. Future developments should focus on enhancing accuracy, developing multimodal AI models, improving empathy through sentiment analysis, and safeguarding against artificial hallucination. Responsible integration requires validation protocols, clear guidelines for appropriate use cases, staff training, and continuous monitoring of AI-assisted outcomes against traditional methods. This balanced approach aims for a symbiotic relationship where ChatGPT augments healthcare providers’ expertise, leading to more efficient, accessible, and high-quality healthcare delivery.
References
Tangsrivimol, J. A., Darzidehkalani, E., Virk, H. U. H., Wang, Z., Egger, J., Wang, M., Hacking, S., Glicksberg, B. S., Strauss, M., & Krittanawong, C. (2025). Benefits, limits, and risks of ChatGPT in medicine. Frontiers in Artificial Intelligence, 8, 1518049. https://doi.org/10.3389/frai.2025.1518049
