Objective: This study aimed to systematically assess the accuracy of ChatGPT-5.2 (OpenAI, USA) in responding to questions on the diagnosis, evaluation, and treatment of pes planus, and to investigate the potential role of artificial intelligence models in orthopaedic clinical practice.

Material and Method: A total of 30 five-option multiple-choice questions covering the diagnosis, clinical evaluation, radiological findings, and treatment approaches of pes planus were prepared based on current literature. The questions were developed through a multi-stage expert evaluation process by three orthopaedic specialists, one of whom had specific experience in foot and ankle surgery, and were subsequently submitted to the ChatGPT-5.2 model. Assuming a 20% probability of random success for five-option questions, accuracy was analyzed using a binomial test, and the 95% confidence interval was calculated using the Wilson method. Differences in accuracy rates across question categories were analyzed using ffisher’s exact test.

Model responses were independently evaluated by three experts, and interobserver agreement was assessed using ffleiss’ kappa coefficient.

Results: The ChatGPT-5.2 model correctly answered 24 of 30 questions, yielding an overall accuracy of 80% (95% CI: 62.7%–90.5%). Under the assumption of a 20% random success rate, the observed accuracy was found to be significant using a one-sided binomial test (p<0.001). Category-based analysis demonstrated high performance in definition, anatomy, and clinical evaluation questions, whereas lower accuracy was observed in treatment and surgical indication items. Inter-rater agreement among experts was at a near-perfect level (ffleiss’ =0.89). Half of the incorrect responses were classified as high clinical risk, all of which were within the treatment/surgical category. ChatGPT-5.2 performance differed significantly across categories (p=0.048).

Conclusion: ChatGPT-5.2 may serve as a useful supplemental tool for orthopaedic information and patient education regarding pes planus; however, it should not replace specialist clinical judgment.

Artificial intelligence ChatGPT Pes planus