BMC Oral Health. 2025 Dec 4. doi: 10.1186/s12903-025-07449-2. Online ahead of print.
ABSTRACT
BACKGROUND: This study aimed to evaluate and compare the quality, reliability, readability, and originality of information provided by two AI-generated chatbots-ChatGPT-4.0 and Google Gemini Pro-regarding primary tooth pulpotomy, a common pediatric dental procedure.
METHODS: Based on the current AAPD guidelines and frequently asked parental inquiries, a total of 20 questions on primary tooth pulpotomy-10 theoretical and 10 clinical-were prepared. Each question was presented to the AI-generated chatbots (ChatGPT-4.0 and Google Gemini Pro) in a new conversation session without providing any guiding prompts. The responses were evaluated using standardized assessment criteria, including DISCERN, EQIP, GQS, FRES, FKRGL, and the iThenticate similarity index. Two experienced pediatric dentists independently assessed the answers after a calibration process to standardize scoring, and inter-rater reliability was confirmed using the intraclass correlation coefficient (ICC).
RESULTS: Gemini Pro demonstrated higher reliability and quality scores (DISCERN and EQIP), while ChatGPT-4 produced responses with higher complexity, requiring university-level reading skills (FKRGL). No significant differences were observed in plagiarism or global quality scores (GQS). Gemini Pro’s responses were more readable, enhancing accessibility for broader audiences.
CONCLUSION: While both AI models generated informative and original content, Gemini Pro provided more reliable and accessible responses, making it a potentially valuable resource for patient and parent education. However, AI-generated information should not replace professional dental consultation. Future AI development should focus on improving source transparency, readability, and clinical relevance. AI-powered language models can be valuable resources for patient and parent education on primary tooth pulpotomy. Gemini Pro’s higher reliability and readability enhance accessibility, while ChatGPT-4’s responses require a higher level of education. These models have the potential to improve access to dental information.
PMID:41340047 | DOI:10.1186/s12903-025-07449-2
