Related Experiment Video
Updated: Jun 28, 2026
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
Artificial Intelligence in Postoperative Sinus Care: A Comparative Study of ChatGPT-4, Google Gemini, and DeepSeek in
Ali M Alsudays1, Khaled A Almanea2, Abdullah A Alhajlah2
1Department of Otolaryngology-Head and Neck Surgery, College of Medicine, King Saud University.
Introduction:
Artificial intelligence (AI) integration into postoperative care has demonstrated significant potential in enhancing patient care and support. This review demonstrates different findings from various studies to evaluate AI's impact on improving postoperative care outcomes, with a specific focus on its application to Functional Endoscopic Sinus Surgery (FESS) in the literature. This study aimed to compare the performance of 3 different large language models in addressing postoperative sinus care questions. The focus is to determining their utility in patient education and support following FESS.
Methodology:
This cross-sectional study was conducted over a 3-month period. Ten standardized questions were adapted from 3 identified online sources (University of Michigan Health, Kevin Caceres, MD, and GhiamMD). Each question was presented to all 3 AI chatbots under identical conditions, generating a total of 30 AI-generated responses for evaluation. A new chat window was used for every question to ensure unbiased responses.
Results:
Findings suggest that the number of words ( P =0.026), number of sentences ( P <0.001), and number of characters per word ( P =0.007) were significantly higher in DeepSeek, but DeepSeek showed significantly lower in the number of words per sentence ( P <0.001). According to evaluators, ChatGPT-4 ratings were better regarding the clarity of responses, whereas DeepSeek ratings were better in completeness. However, Google Gemini performed the least among the AI Chatbots. Interestingly, reading difficulties in the responses from Google Gemini and DeepSeek were somewhat higher than ChatGPT-4.
Conclusion:
Both ChatGPT-4 and DeepSeek had comparable ratings on the response's accuracy, relevance, and usefulness. However, the number of words, sentences, and characters per word was significantly higher in DeepSeek. Interestingly, Google Gemini's ratings lagged behind both ChatGPT-4 and DeepSeek. Further investigations are required to determine which AI Chatbots offer the best responses in various clinical case scenarios, particularly in postoperative care in our region.
