Assessment of ChatGPT-Generated Endodontic Patient Education Materi als: A Readability and Accuracy Analysis

Abstract

Background: The increasing use of artificial intelligence–based conversational models  such as ChatGPT has transformed the way patients access healthcare-related information.  In endodontics, patients frequently seek online educational material regarding root canal  treatment and related conditions. However, the readability and scientific accuracy of AI generated dental information remain uncertain. Aim: To evaluate the readability and scientific accuracy of ChatGPT-generated patient  education materials related to common endodontic conditions and procedures. Materials and Methods: A cross-sectional analytical study was conducted using 100 patient oriented responses generated by ChatGPT. Ten standardized endodontic prompts related  to root canal treatment, dental abscess, infected teeth, post-treatment care, and treatment  complications were entered ten separate times under identical settings. The generated  responses were collected without modification. Readability assessment was performed using  the Flesch Reading Ease Score, Flesch–Kincaid Grade Level, SMOG Index, and Gunning  Fog Index. Scientific accuracy and quality of information were independently evaluated by  three experienced endodontists using a structured 5-point Likert scale assessing accuracy,  completeness, clarity, misleading information, and clinical usefulness. Statistical analysis was  performed using descriptive statistics and Cohen’s kappa test for inter-examiner reliability. Results: ChatGPT-generated educational materials demonstrated moderate to high  readability, with most responses corresponding to a secondary school reading level. The  majority of responses showed good scientific accuracy, clarity, and clinical usefulness.  However, certain responses lacked comprehensive explanations regarding procedural  complications and emergency management. Minor inconsistencies and oversimplified  information were observed in selected outputs. Inter-examiner agreement among evaluators  was high. Conclusion: ChatGPT demonstrated promising potential in generating readable and  reasonably accurate endodontic patient education materials. Nevertheless, professional  verification by dental experts remains necessary before clinical use to ensure reliability and  prevent dissemination of incomplete or misleading information.