İşisağ Özer, Karakaya Kevser
Faculty of Dentistry Department of Prosthodontics, Afyonkarahisar Health Sciences University, Güvenevler Neighborhood, İnönü Boulevard No:4,, Afyonkarahisar, Turkey.
Clin Oral Investig. 2025 Sep 2;29(9):433. doi: 10.1007/s00784-025-06521-z.
The aim of this study is to evaluate the effectiveness of ChatGPT 4o in planning tooth-supported fixed prostheses by examining the accuracy, repeatability, and consistency of its responses across various prosthodontic cases.
1,140 responses were generated for 38 questions regarding simple and complex prosthetic cases, using case schemas from Fundamentals of Fixed Prosthodontics. Each question was asked 30 times at different times of day (morning, afternoon, and evening) to assess response consistency.
ChatGPT demonstrated high accuracy (97.08%) for simple fixed prostheses but struggled with complex cases, showing a low accuracy rate of 19.69%. Statistically significant differences were observed in the response accuracy between question categories, indicating that ChatGPT performs better with straightforward cases than with complex ones. Repeatability was high across all question types, though accuracy varied, especially in complex scenarios.
The findings highlight ChatGPT's potential as an assistive tool in simple prosthodontic cases yet emphasize the need for clinician expertise in more complex treatment planning.
While ChatGPT shows promise as a supplementary tool for dental education and practice, reliance on AI alone for intricate cases remains premature.
本研究旨在通过检查ChatGPT 4o在各种修复病例中的回答准确性、可重复性和一致性，评估其在设计牙支持固定修复体方面的有效性。
使用《固定义齿修复学基础》中的病例模式，针对简单和复杂修复病例的38个问题生成了1140个回答。每个问题在一天中的不同时间（上午、下午和晚上）被询问30次，以评估回答的一致性。
ChatGPT在简单固定修复体方面表现出较高的准确性（97.08%），但在复杂病例中表现不佳，准确率仅为19.69%。不同问题类别之间的回答准确性存在统计学上的显著差异，这表明ChatGPT在简单病例中的表现优于复杂病例。尽管准确性有所不同，尤其是在复杂场景中，但所有问题类型的可重复性都很高。
研究结果凸显了ChatGPT在简单修复病例中作为辅助工具的潜力，但也强调了在更复杂的治疗计划中临床医生专业知识的必要性。
虽然ChatGPT有望成为牙科教育和实践的辅助工具，但仅依靠人工智能处理复杂病例仍为时过早。
