Can Artificial Intelligence Chatbots Guide Preclinical Endodontic Training?

This study aims to evaluate the accuracy of responses from four artificial intelligence (AI) chatbots regarding endodontic access cavity, root canal preparation, and irrigation in preclinical endodontic training to assess their potential as educational tools. A total of 20 questions were posed to four AI chatbots (ChatGPT-3.5, DeepSeek-V3, Gemini 2.0 Flash, and Microsoft Copilot) three times daily over a 10-day period. The questions were developed based on guidelines from the British Endodontic Society, the Australian Society of Endodontology, and the European Society of Endodontology. Responses were compared with these guidelines and categorized as correct, incorrect, or insufficient. Pearson Chi-Square tests were performed to assess consistency and accuracy of responses (p < 0.05). A total of 2,400 responses from the four AI chatbots were analyzed. DeepSeek-V3 achieved the highest accuracy with 94.3% correct responses, followed by ChatGPT-3.5 with 89.0%, Gemini 2.0 Flash with 88.7%, and Microsoft Copilot with 79.2%. DeepSeek-V3 demonstrated significantly better performance than ChatGPT-3.5, Gemini 2.0 Flash, and Microsoft Copilot (p < 0.05), while ChatGPT-3.5 and Gemini 2.0 Flash exhibited similar accuracy (p > 0.05), and Microsoft Copilot showed significantly lower accuracy than the other platforms (p < 0.05). AI chatbots show promise as educational tools in preclinical endodontic training, though accuracy varies across chatbots. Further improvements are needed, especially in endodontics, to enhance performance. These findings highlight the need to improve AI for dental education.