Evaluating ChatGPT-4o's Quality and Readability in Preoperative Facial Plastic Surgery Counseling: A Vignette-Based Analysis.

[BACKGROUND] Generative artificial intelligence (AI) models such as ChatGPT have demonstrated potential in medical education and patient communication and education. However, their utility for delivering personalized preoperative information in plastic surgery remains unsettled.

[METHODS] This study evaluated ChatGPT-4o's ability to generate preoperative counseling for the 5 most common aesthetic procedures, using vignette-style prompts featuring a standardized patient profile. Responses were assessed for quality using the DISCERN instrument, specificity via a Likert scale, and readability through the Flesch-Kincaid Grade Level score.

[RESULTS] ChatGPT's responses achieved moderate overall quality. DISCERN total scores ranged from 40 to 44.7 out of 75 across procedures. Reliability was high for clarity and relevance but low for source transparency and encouragement of shared decision-making. Specificity scores were low to moderate, with the highest specificity observed for abdominoplasty-related responses (mean Likert score 2.6). Readability remained within the recommended range, with an average Flesch-Kincaid Grade Level of 8.4, suitable for most patient populations.

[CONCLUSION] ChatGPT-4o can generate preoperative educational materials at an appropriate reading level with moderate quality and specificity. However, its lack of source disclosure, minimal emphasis on alternative treatments, and insufficient promotion of physician consultation highlight its limitations. ChatGPT may serve as an adjunct to physician counseling but is not a substitute for personalized medical consultation by a plastic surgeon.