Assessing the Usefulness and Quality of Artificial Intelligence-Generated Content on Bone Graft Materials in Dentistry Based on Patient Questions

Background: This study evaluated the performance of three Artificial Intelligence (AI) conversational agents-ChatGPT-4, DeepSeek and Copilot-in providing clinically relevant information on bone graft materials in dentistry. Methodology: A standardized set of questions related to bone graft types, indications, effectiveness, risks and patient comfort was posed to each model and these questions were derived from commonly asked patient inquiries. Responses were independently assessed by calibrated reviewers using validated tools: CLEAR criteria, a modified Global Quality Score (mGQS), a 5-point Likert scale for accuracy, a 4-point usefulness scale and readability metrics (Flesch Reading Ease and Flesch-Kincaid Grade Level). Results: ChatGPT-4 outperformed Copilot in CLEAR scores (p = 0.030) and exceeded both DeepSeek and Copilot in mGQS (p = 0.022 and p = 0.017, respectively). However, no significant differences were observed in accuracy or readability (p > 0.05). ChatGPT-4 scored significantly lower in usefulness compared to DeepSeek and Copilot (p < 0.05). Negative correlations were found between mGQS and usefulness (r = -0.807) and between readability and grade level (r = -0.938). Conclusion: ChatGPT-4 provided higher-quality and more comprehensive answers, but all models showed limitations in readability and usefulness. Improving the clarity and practical relevance of AI-generated content is essential to support dental education and enhance patient communication.