Emrah ISIKTEKIN
Turkish Journal of Plastic Surgery - 2026;34(4):146-151
Background: Artificial intelligence (AI)-based language models are increasingly used by patients to obtain medical information; however, the quality and reliability of their outputs in esthetic surgery remain unclear. Objective: The objective of this study was to comparatively evaluate the quality, completeness, and reliability of AI-generated responses to frequently asked questions regarding facelift surgery using a phase-based expert assessment. Materials and Methods: Fifteen frequently asked questions related to facelift surgery were selected from the American Society of Plastic Surgeons and categorized into preoperative, intraoperative, and postoperative phases. Responses were generated using ChatGPT-4, Gemini 2.0 Flash, and Claude 3.5 Sonnet. Ten European board-certified plastic surgeons independently evaluated all responses in a blinded manner using the Global Quality Score (GQS) and the Quality Analysis of Medical AI (QAMAI) tool. Statistical analyses were performed using repeated measures methods, with significance set at P < 0.05. Results: Overall GQS values did not differ significantly among models (P = 0.135), whereas QAMAI scores showed a significant difference (P = 0.007), with Gemini 2.0 Flash and Claude 3.5 Sonnet outperforming ChatGPT-4. Phase-based analysis demonstrated significantly higher QAMAI scores for Gemini 2.0 Flash and Claude 3.5 Sonnet across all phases (preoperative P = 0.022. intraoperative P = 0.037, and postoperative P = 0.008). Post hoc analysis revealed a significant difference between Claude 3.5 Sonnet and ChatGPT-4 in the postoperative phase (P = 0.018). Interrater reliability demonstrated good agreement at the panel level (GQS intraclass correlation coefficient [ICC] =0.749 and QAMAI ICC = 0.813). Conclusions: AI language models can generate generally acceptable information on facelift surgery; however, their reliability varies. Gemini 2.0 Flash and Claude 3.5 Sonnet demonstrated superior performance compared to ChatGPT-4, highlighting the need for cautious use of AI-generated medical information and continued clinician oversight in esthetic surgery patient education.
