Artificial Intelligence as a Partner in Ankylosing Spondylitis Care: Evaluating ChatGPT’s Role and Performance

Aim: Artificial Intelligence may have significant potential to assist clinicians in decision-making and diagnosis, especially in units dependent on up-to-date guidelines such as rheumatology. This study aims to evaluate the effectiveness of ChatGPT in providing clinicians with evidence-based information about ankylosing spondylitis (AS). Material and Method: Frequently asked questions (FAQs) about AS were developed by reviewing commonly accessed patient-oriented websites, social media platforms, and official hospital pages. Questions were designed based on scientific guidelines, particularly the American College of Rheumatology (ACR) and Assessment of SpondyloArthritis international Society (ASAS)-European League Against Rheumatism (EULAR) axial spondyloarthritis guidelines. ChatGPT's responses were evaluated on a 1-to-4 scale. Each question was posed twice to assess reproducibility, with consistency defined by identical scores across both attempts. Results: ChatGPT demonstrated an overall accuracy of 81.9% in its responses to 72 FAQs. The highest accuracy (91.7%) was observed in responses related to the prevention of AS. Of the 36 questions based on ACR and ASAS-EULAR guidelines, ChatGPT provided accurate answers for 22 (61.1%), with three responses receiving the lowest grade (4). Reproducibility of ChatGPT's responses was 88.8% across all FAQs and 83.3% for guideline-specific questions. Conclusion: This study highlights the potential of ChatGPT as a supportive tool for patient education and clinician reference, particularly for general FAQs. However, accuracy for questions derived from ACR and ASAS-EULAR guidelines was lower (61.1%), emphasizing the need for clinician oversight.