Artificial Intelligence in Periodontology: Performance Evaluation of ChatGPT, Claude, and Gemini on the In-service Examination

Abstract

ABSTRACT Background Artificial intelligence (AI) language models have shown potential as educational tools in healthcare, but their accuracy and reliability in periodontology education require further evaluation. In this study we aimed to assess and compare the performance of three prominent AI language models—ChatGPT-4o, Claude 3 Opus, and Gemini Advanced—with second-year periodontics residents across the United States on the American Academy of Periodontology 2024 in-service examination. Methods We conducted a cross-sectional study using 331 multiple-choice questions from the 2024 periodontology in-service examination. We evaluated and compared the performances of ChatGPT-4o, Claude 3 Opus, and Gemini Advanced across various question domains. The results of second-year periodontics residents served as a benchmark. Results ChatGPT-4o, Gemini Advanced, and Claude 3 Opus significantly outperformed second-year periodontics residents across the United States, with accuracy rates of 92.7 percent, 81.6 percent, and 78.5 percent, respectively, compared to the residents’ 61.9 percent. The differences in performance among the AI models were statistically significant ( p < 0.001). Percentile rankings underscored the superior performance of the AI models, with ChatGPT-4o, Gemini Advanced, and Claude 3 Opus placing in the 99.95th, 98th, and 95th percentiles, respectively. Conclusion ChatGPT-4o displayed superior performance compared to Claude 3 Opus and Gemini Advanced. The results highlight the potential of AI large language models (LLMs) as educational tools in periodontology and emphasize the need for ongoing evaluation and validation as these technologies evolve. Researchers should explore both the integration of AI language models into periodontal education and their impact on learning outcomes and clinical decision-making.

Ähnliche Arbeiten