Related Experiment Video
Updated: Jun 12, 2025
07:32
Author Spotlight: 3D Movement Assessment of Maxillary Posterior Teeth in Clear Aligner Treatment
Published on: February 23, 2024
1.0K
A Comparative Analysis of Three Large Language Models on Bruxism Knowledge
Elisa Souza Camargo1, Isabella Christina Costa Quadras1, Roberto Ramos Garanhani2
1Graduate Program in Dentistry, Pontifícia Universidade Católica do Paraná, Curitiba, Brazil.
Journal of Oral Rehabilitation
|February 6, 2025
Summary
Large language models (LLMs) show moderate accuracy and high consistency in answering bruxism questions. While Gemini offers better readability, AI tools should not replace professional dental advice for bruxism.
Area of Science:
- Medical Informatics
- Artificial Intelligence in Healthcare
Background:
- Artificial Intelligence (AI) is increasingly used in health research.
- The effectiveness of Large Language Models (LLMs) for bruxism information is unevaluated.
Purpose of the Study:
- Assess LLM readability, accuracy, and consistency for bruxism FAQs.
- Evaluate ChatGPT-3.5, ChatGPT-4, and Gemini performance.
Main Methods:
- Identified top bruxism topics using Google Trends.
- Selected 30 FAQs, queried three LLMs twice (T1, T2).
- Measured readability (FRE, FKG), accuracy (3-point scale), and consistency (T1 vs. T2).
Main Results:
- No significant difference in Flesch Reading Ease (FRE) across models.
- Gemini had lower Flesch-Kincaid Grade Level (FKG) scores.
- Average accuracy: GPT-3.5 (68.33%), GPT-4 (65%), Gemini (55%).
- All models showed substantial consistency (GPT-3.5 highest at 95%).
Conclusions:
- Gemini's responses may be more accessible to patients.
- LLMs offer moderate accuracy and high consistency for bruxism information.
- AI tools are supplementary and should not replace professional dental guidance.
