Artificial intelligence applications in tooth avulsion: comparative accuracy of ChatGPT and DeepSeek

Background: The accuracy and performance of artificial intelligence (AI)-based chatbots in clinical applications can directly influence healthcare outcomes. In cases of dental trauma, adherence to the International Association of Dental Traumatology (IADT) guidelines is essential for clinical success. Although the use of AI in healthcare is increasing, few studies have evaluated the ability of chatbots to provide accurate information in dental trauma. This study aimed to evaluate and compare the performance of the ChatGPT and DeepSeek platforms in providing guideline-based information on the management of dental avulsion, using the IADT guidelines as a reference standard. Methods: Based on the IADT guidelines, 25 questions (12 yes/no and 13 open-ended) were posed to ChatGPT-3.5 and DeepSeek over the course of one week. Two independent researchers asked each question three times daily. Responses were classified as correct, incorrect, or insufficient according to the guidelines. Statistical analyses were conducted to assess agreement and accuracy. Results: A total of 1050 responses were analyzed. DeepSeek demonstrated moderate agreement with the guideline-based answers (κ ≈ 0.52; 95% confidence interval (CI): 0.48–0.55; p < 0.001), whereas ChatGPT showed weak-to-moderate agreement (κ ≈ 0.44; 95% CI: 0.40–0.48; p < 0.001). The mean accuracy difference between the two platforms was approximately 7% (p = 0.001). Conclusions: ChatGPT and DeepSeek have potential as knowledge resources for healthcare applications. However, their accuracy and consistency in addressing dental avulsion-related questions remain limited. Clinicians should consider these systems as complementary tools that support, but do not replace, clinical expertise and decision-making. Further research should explore AI models specifically trained in dental trauma to determine their clinical utility.