Evaluation of Responses Provided by Artificial Intelligence Chatbots to Questions on Hyperbaric Oxygen Therapy

Abstract

Background: The aim of this study was to evaluate the accuracy, reliability, and readability of responses provided by artificial intelligence chatbot systems to questions related to hyperbaric oxygen therapy, and to determine their potential as a source of medical information.Methods: The 20 most frequently searched open-ended questions on HBOT were identified using Google Trends and directed to ChatGPT 5.2, ChatGPT-4.0, ChatGPT-4.1, Copilot, Gemini, and Perplexity. All responses were independently evaluated by two expert physicians. Accuracy and completeness were assessed using Likert scales, reliability using the mDISCERN tool, and readability using the SMOG index. Statistical analyses included the Kruskal–Wallis, Dunn, Mann–Whitney U, and Fisher’s exact tests.Results: According to median values, no significant differences were observed among ChatGPT 5.2, ChatGPT 4.0, ChatGPT 4.1, Gemini, and Copilot in terms of accuracy; however, Perplexity demonstrated significantly lower scores (p