Author ORCID Identifier

https://orcid.org/0009-0000-9638-929X

Semester

Summer

Date of Graduation

2025

Document Type

Thesis

Degree Type

MS

College

School of Dentistry

Committee Chair

Samuel Dorn, DDS

Committee Member

R. Constance Wiener, D.M.D, PhD, MA

Committee Member

Steven B. Whitaker, D.D.S.

Abstract

Comparative Analysis of Safety, Accuracy, Comprehensiveness, and Organization in Endodontic Patient Education Material: American Association of Endodontics (AAE) vs. AI-Generated Content. Fabiola Salgado Reyes, DMD Introduction: With patients increasingly using AI chatbots like ChatGPT4o and Copilot, it is essential to assess whether these tools can effectively support patient understanding of complex endodontic procedures. AI-generated responses may occasionally provide inaccurate, unsafe, disorganized, or content that is not comprehended. The purpose of this study is to examine if dental experts and patients can determine if patient education materials (PEMS) were 1) generated by AI; 2) medically safe; 3) accurate; 4) can be comprehended; and 5) organized. Secondarily, the PEMS were evaluated for their readability. Methods: AAE PEMs were compared to two AI-generated PEMs (ChatGPT4o and Copilot Premium accessed via subscription from Nov 13–Dec 14, 2024) by dental experts and patients. Content included information on diagnosis, symptoms, risk factors, treatment, and prognosis. Each dental expert evaluated all three PEM. Each patient receiving endodontic therapy at West Virginia School of Dentistry who agreed to participate evaluated one randomized PEM. Both groups evaluated the PEMs for safety, accuracy, comprehensibility, and organization (0–4 scale). They also determined if each PEM was AI generated or human generated. PEMs were also evaluated for readability. Results: Dental experts did not distinguish a quality difference in safety, accuracy, organization, or comprehensibility among the 3 PEM (p>0.05). Patients could not distinguish between AI and human generated PEM (p=0.666). However, Copilot PEM consistently was scored higher by patients in safety, accuracy, and comprehension (p < 0.05). Conclusion: AI-generated PEMs, especially from Copilot, demonstrate strong potential as accurate, readable, and well-received alternatives to traditional materials, supporting their integration into clinical patient education.

Download

Share

COinS