Evaluating the accuracy and reproducibility of ChatGPT responses in the context of cochlear implantation

Background The use of ChatGPT in the field of otolaryngology is increasing; however, there are not enough studies related to cochlear implants. Aim/objectives To assess the accuracy and reproducibility of ChatGPT (GPT-4o) responses to questions about cochlear implantation, evaluating its potential role in patient education. Material and methods A total of 104 questions across five categories (basic, preoperative, surgical, postoperative care, postoperative expectations) were selected from reliable online sources. Each was posed twice to ChatGPT-4o in separate sessions. Responses were graded by two reviewers for accuracy (comprehensive/correct, partially correct, misleading, or incorrect/irrelevant). Reproducibility was assessed based on consistency across sessions. Discrepancies were resolved by a third expert reviewer. Results Of 104 responses, 84.6% were correct, 5.8% partially correct, 6.7% misleading, and 2.9% incorrect. Reproducibility was 88.4% overall and 100% in the surgery category. No significant differences were found between question categories for accuracy (p = 0.829) or reproducibility (p = 0.348). Conclusion and significance ChatGPT provided highly accurate and reproducible responses to cochlear implant-related questions, supporting its use as an educational tool. Nonetheless, expert review remains essential for complex or critical topics.