S1169 Artificial Intelligence-Based Patient Education for Irritable Bowel Syndrome: A Comparative Evaluation of ChatGPT and Gemini

Introduction: Irritable bowel syndrome (IBS) is a prevalent functional gastrointestinal disorder that significantly affects patients' quality of life. Despite advancements in diagnosis and treatment, patient education remains a critical gap in care. The rise of large language models (LLMs) such as ChatGPT and Gemini presents a novel opportunity to enhance patient understanding through accessible and scalable digital communication. This study aims to assess the accuracy, comprehensiveness, readability, and empathy of these AI models in delivering educational content about IBS. Methods: A total of 39 frequently asked questions related to IBS were compiled from major gastrointestinal websites and categorized into 6 domains: general understanding, symptoms and diagnosis, causes and risk factors, dietary considerations, treatment and management, and lifestyle factors. Responses from ChatGPT-4 and Gemini-1 were generated using identical prompts. Two experienced gastroenterologists independently evaluated each response for comprehensiveness and factual accuracy. A third reviewer resolved any disagreements. Readability was assessed using established indices such as Flesch Reading Ease and SMOG Index. Empathy was rated using a four-point Likert scale by 3 reviewers. Descriptive statistics and t-tests were used for analysis. Results: Both ChatGPT and Gemini provided generally informative answers. Gemini tended to produce slightly more comprehensive and readable content, while ChatGPT was consistently rated as more empathetic. However, both models generated content at a reading level beyond the average patient, potentially limiting accessibility. Neither model exhibited a high rate of factual inaccuracies, but occasional mixed-quality responses highlighted the need for expert oversight. Conclusion: LLMs such as ChatGPT and Gemini represent promising tools for augmenting patient education in gastroenterology. Each model demonstrated strengths: Gemini in factual comprehensiveness and readability and ChatGPT in empathetic delivery. However, the use of complex language remains a barrier to broader applicability. To be effective educational adjuncts, these tools must evolve to deliver accurate, emotionally intelligent, and easily understood content. Clinical oversight and tailored communication are necessary to ensure equitable access to health information.