Introduction: Accurate age estimation is essential in medical and forensic practice. Dental development is among the most dependable biological indicators, and radiographic methods such as the Cameriere method have been validated across populations. Recently, vision-enabled large language models, including Chat Generative Pretrained Transformer-5 (ChatGPT-5), have attracted attention for image analysis. This study evaluated the performance of ChatGPT-5 in dental age (DA) estimation and compared its agreement with chronological age (CA) with that of the Cameriere method.
Materials and Methods: This retrospective, comparative, methodological study analyzed 116 cropped panoramic radiographs of the mandibular left region from Turkish children aged 4.0–13.99 years. DA was estimated digitally using ImageJ software by two calibrated pediatric dentists applying the Cameriere method, and by ChatGPT-5 under two standardized prompting conditions (unguided and Cameriere-guided). Analyses were performed on the overall sample without sex-specific or age-specific subgroup evaluations. Agreement with CA was assessed using mean absolute error (MAE) and root mean square error (RMSE). Paired comparisons were conducted using paired t-tests or Wilcoxon signed-rank tests, depending on data distribution. Reliability was evaluated using intraclass correlation coefficients (ICC). Results: The Cameriere method demonstrated the highest accuracy and reliability (MAE=0.63 years; RMSE=0.81 years). ChatGPT-5 produced estimates that have greater variation. Performance improved when guided by the Cameriere formula, but reliability remained moderate (ICC=0.57).
Discussion and Conclusion: While the Cameriere method provided more consistent age estimations, ChatGPT-5’s estimates were more variable and insufficiently precise for clinical or forensic use.