Comparative Evaluation of Information Quality on Colon Cancer for Patients: A Study of ChatGPT-4 and Google

Abstract

Introduction This study aimed to evaluate and compare the quality and reliability of information provided by two widely used digital platforms, ChatGPT-4 and Google, on frequently asked questions about colon cancer. With the growing popularity of these platforms, individuals increasingly turn to them for accessible health information, yet questions remain regarding the accuracy and reliability of such content. Given that colon cancer is a prevalent and serious condition, trustworthy information is essential to support patient education, facilitate informed decision-making, and potentially improve patient outcomes. Therefore, the objective was to determine which platform offers more reliable and accurate medical information on colon cancer, using established evaluation criteria to assess the quality of information. Methods Twenty frequently asked questions about colon cancer were selected based on search popularity and relevance to patients and then searched using ChatGPT-4 and Google. Responses were evaluated using tools such as DISCERN (reliability), Global Quality Score (GQS), Journal of the American Medical Association (JAMA) criteria (accuracy), SAM (suitability), Flesch-Kincaid Readability Test, HITS (user experience), and VPI (visibility). Statistical analyses determined significant differences between the platforms (p < 0.05). ChatGPT-4 scored significantly higher than Google on DISCERN, GQS, and JAMA, indicating greater reliability, accuracy, and comprehensibility (p < 0.001). Both platforms showed similar readability scores, but ChatGPT-4 rated higher for patient suitability (SAM, p < 0.01) and user-friendliness (HITS, p < 0.01). Although Google exhibited higher visibility (VPI), the limited HONcode certification raised concerns about the reliability of its results. Results ChatGPT-4 scored significantly higher than Google on DISCERN, GQS, and JAMA criteria, demonstrating superior reliability, accuracy, and comprehensibility (p < 0.001). While both platforms had comparable readability scores on the Flesch-Kincaid Readability Test, ChatGPT-4 was rated as more suitable for patient education according to SAM criteria (p < 0.01). Furthermore, ChatGPT-4 was found to be more user-friendly and offered more structured information based on the HITS scale (p < 0.01). Although Google showed higher visibility according to the VPI, the limited presence of HONcode-certified results raised concerns about the reliability of its information. Conclusion ChatGPT-4 proved to be a more reliable and higher-quality source of medical information compared to Google, particularly for patient queries about colon cancer. AI-based platforms such as ChatGPT-4 hold promise for enhancing patient education and providing accurate medical information, although further research is needed to confirm these findings across different medical topics and larger populations.