Dentist-Led Teledentistry Outperformed ChatGPT in Caries Detection Study

Edited by Hygienetown staff

A dentist assessing smartphone photographs remotely detected caries more accurately than ChatGPT analyzing the same images, according to a preliminary diagnostic accuracy study published Aug. 22 in BMC Oral Health.

Researchers evaluated 40 tooth images, a sample size determined by power analysis. Standardized smartphone photographs were assessed independently two ways, once through ChatGPT-based analysis and once by dental professionals working remotely. Both sets of results were compared against direct clinical examination, which served as the reference standard.

Teledentistry scored higher on every measure. It reached 89.5% sensitivity against 78.9% for ChatGPT, 85.7% specificity against 66.7%, and 87.5% overall accuracy against 72.5%. Positive predictive value was 85% versus 68.2%, and negative predictive value 90% versus 77.8%.

Discriminative ability followed the same pattern. Teledentistry recorded an area under the curve of 0.876, compared with 0.728 for ChatGPT.

Neither method differed significantly from the clinical reference standard when tested with McNemar’s test, but the investigators reported that ChatGPT showed greater inconsistency. Agreement with clinical examination was assessed using Cohen’s kappa.

The authors concluded that teledentistry demonstrated higher accuracy and stronger agreement with clinical examination than ChatGPT, and may serve as a reliable diagnostic adjunct in remote settings. They wrote that general-purpose AI tools need further refinement and validation, and pointed toward larger datasets and better image quality as priorities.

Inas Karawia, Aya Mahmoud, Rewan Ali, Marie Kamal, Aya Gaber, and Mohamed Ashraf Hall conducted the study.

Forty images is a small sample, and the design compared each method against clinical examination rather than testing the two methods directly against each other, so the gap between them was not subjected to a statistical comparison. Radiographs were not part of the reference standard, which limits what the accuracy figures can say about lesions not visible on a photograph.

Sources:
BMC Oral Health, “Diagnostic accuracy of ChatGPT and teledentistry compared with dentist examination in detecting dental caries: preliminary diagnostic accuracy study,” published online Aug. 22, 2026: link.springer.com/10.1186/s12903-026-09638-z