Abstract
Background The increasing reliance on the internet for health-related information has driven interest in artificial intelligence (AI) applications in healthcare. ChatGPT has demonstrated strong performance in medical exams, raising questions about its potential use in patient education. However, no prior study has evaluated the reliability of ChatGPT in explaining medical terms. This study investigates whether ChatGPT-4 is a reliable tool for translating frequently used medical terms into language that patients can understand. Methodology A total of 105 frequently used medical terms were selected from the University of San Diego's medical terminology list. Four groups - general practitioners, resident physicians, specialist physicians, and ChatGPT-4 - were tasked with defining these terms. Responses were classified as correct or incorrect. Statistical analyses, including chi-square and post-hoc tests, were conducted to compare accuracy rates across groups. Results ChatGPT-4 achieved a 100% accuracy rate, outperforming specialist physicians (98.1%), resident physicians (93.3%), and general practitioners (84.8%). The differences in accuracy rates between groups were statistically significant (χ²=25.99, p<0.00001). Post-hoc analyses confirmed significant pairwise differences, such as ChatGPT-4 vs. specialist physicians (p<0.001) and specialist physicians vs. resident physicians (p=0.02). Conclusions ChatGPT-4 demonstrated superior reliability in translating medical terms into understandable language, surpassing even highly experienced physicians. These findings suggest that ChatGPT could be a valuable auxiliary tool for improving patient comprehension of medical terminology. Nonetheless, the importance of consulting healthcare professionals for clinical decision-making remains crucial.