Readability, reliability and quality of responses generated by ChatGPT, gemini, and perplexity for the most frequently asked questions about pain

It is clear that artificial intelligence-based chatbots will be popular applications in the field of healthcare in the near future. It is known that more than 30% of the world's population suffers from chronic pain and individuals try to access the health information they need through online platforms before applying to the hospital. This study aimed to examine the readability, reliability and quality of the responses given by 3 different artificial intelligence chatbots (ChatGPT, Gemini and Perplexity) to frequently asked questions about pain. In this study, the 25 most frequently used keywords related to pain were determined using Google Trend and asked to every 3 artificial intelligence chatbots. The readability of the response texts was determined by Flesch Reading Ease Score (FRES), Simple Measure of Gobbledygook, Gunning Fog and Flesch-Kincaid Grade Level readability scoring. Reliability assessment was determined by the Journal of American Medical Association (JAMA), DISCERN scales. Global Quality Score and Ensuring Quality Information for Patients (EQIP) score were used in quality assessment. As a result of Google Trend search, the first 3 keywords were determined as "back pain," "stomach pain," and "chest pain." The readability of the answers given by all 3 artificial intelligence applications was determined to be higher than the recommended 6th grade readability level (P < .001). In the readability evaluation, the order from easy to difficult was determined as Google Gemini, ChatGPT and Perplexity. Higher GQS scores (P = .008) were detected in Gemini compared to other chatbots. Perplexity had higher JAMA, DISCERN and EQIP scores compared to other chatbots, respectively (P < .001, P < .001, P < .05). It has been determined that the answers given by ChatGPT, Gemini, and Perplexity to pain-related questions are difficult to read and their reliability and quality are low. It can be stated that these artificial intelligence chatbots cannot replace a comprehensive medical consultation. In artificial intelligence applications, it may be recommended to facilitate the readability of text content, create texts containing reliable references, and control them by a supervisory expert team.