How accurate and useful are AI chatbots in interpreting pediatric intraoral photographs?
- 1Faculty of Dentistry, Department of Pediatric Dentistry, Lokman Hekim University, Ankara, Turkey
- 2Faculty of Dentistry, Department of Pediatric Dentistry, Antalya Bilim University, Antalya, Turkey
Abstract
Background: Accurate recognition of oral conditions is essential for parents to seek timely dental care for their children. However, many parents struggle to identify common pediatric dental problems. Artificial intelligence (AI) tools capable of interpreting clinical images may help bridge this knowledge gap by providing understandable and actionable guidance. This study assessed the diagnostic accuracy, consistency, and communication quality of AI-generated responses to intraoral photographs of pediatric dental conditions.
Methods: Three large language models —ChatGPT-4o, ChatGPT-5, and Gemini 2.0 Flash—were tested using standardized intraoral photographs of common pediatric dental conditions. Each model was prompted with the same diagnostic question (What is this condition, and what should I do? ). Responses were evaluated for diagnostic accuracy, understandability, actionability, and overall quality using the Patient Education Materials Assessment Tool (PEMAT) and the Global Quality Scale (GQS). Statistical comparisons were performed using the Kruskal–Wallis test (p < 0.05).
Results: ChatGPT-4o achieved the highest diagnostic accuracy (81.48%), followed by ChatGPT-5 (75.93%) and Gemini 2.0 Flash (35.19%) (p < 0.001). Significant differences were also observed in PEMAT and GQS scores (p < 0.001).
Conclusions: ChatGPT-4o and ChatGPT-5 demonstrated moderate-to-high accuracy, yet variability in communication quality persists. These tools may assist parent education but cannot replace professional evaluation.
Related
- ArchCaries-YOLO26: a dentition-aware framework for automated detection of primary and permanent tooth caries in intraoral photographs
- Extraoral Versus Intraoral Photobiomodulation for the Prevention of Oral Mucositis in Pediatric Patients Undergoing High-Dose Methotrexate Chemotherapy.
- Comparable efficacy of extraoral and intraoral photobiomodulation in preventing oral mucositis during high-dose methotrexate therapy in pediatric oncology: A single-blind randomized controlled trial.
- Operators matter \textendash An assessment of the expectations, perceptions, and performance of dentists, postgraduate students, and dental prosthetist students using intraoral scanning
- Development and Evaluation of an Offline Vision-Language Model System for Estimating Nutrients from Meal Photographs on Personal Devices: The PLANS (Personal-device Local Assessment of Nutrients System) Study
BibTeX
@article{ozdemir2026accurate,
title = {{How accurate and useful are AI chatbots in interpreting pediatric intraoral photographs?}},
author = {Merve Ozdemir and Dilan Altun},
journal = {BMC Oral Health},
year = {2026},
month = aug,
doi = {10.1186/s12903-026-09554-2},
url = {https://doi.org/10.1186/s12903-026-09554-2},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - How accurate and useful are AI chatbots in interpreting pediatric intraoral photographs?
AU - Merve Ozdemir
AU - Dilan Altun
PY - 2026
DA - 2026/08/08
JO - BMC Oral Health
DO - 10.1186/s12903-026-09554-2
UR - https://doi.org/10.1186/s12903-026-09554-2
AB - Background: Accurate recognition of oral conditions is essential for parents to seek timely dental care for their children. However, many parents struggle to identify common pediatric dental problems. Artificial intelligence (AI) tools capable of interpreting clinical images may help bridge this knowledge gap by providing understandable and actionable guidance. This study assessed the diagnostic accuracy, consistency, and communication quality of AI-generated responses to intraoral photographs of pediatric dental conditions. Methods: Three large language models —ChatGPT-4o, ChatGPT-5, and Gemini 2.0 Flash—were tested using standardized intraoral photographs of common pediatric dental conditions. Each model was prompted with the same diagnostic question (What is this condition, and what should I do? ). Responses were evaluated for diagnostic accuracy, understandability, actionability, and overall quality using the Patient Education Materials Assessment Tool (PEMAT) and the Global Quality Scale (GQS). Statistical comparisons were performed using the Kruskal–Wallis test (p < 0.05). Results: ChatGPT-4o achieved the highest diagnostic accuracy (81.48%), followed by ChatGPT-5 (75.93%) and Gemini 2.0 Flash (35.19%) (p < 0.001). Significant differences were also observed in PEMAT and GQS scores (p < 0.001). Conclusions: ChatGPT-4o and ChatGPT-5 demonstrated moderate-to-high accuracy, yet variability in communication quality persists. These tools may assist parent education but cannot replace professional evaluation.
SN - 1472-6831
ER -