Karnan Nithin, Nair Sruthi, Fidai Farhaan Firoz, Gurrala Sri Vidhya, Salim Jasmine, Gomma Ahmed
Internal Medicine, KAP Viswanathan Government Medical College, Tiruchirappalli, IND.
Geriatrics, Birmingham Heartlands Hospital, Birmingham, GBR.
Cureus. 2025 Apr 10;17(4):e81993. doi: 10.7759/cureus.81993. eCollection 2025 Apr.
Diabetes management involves using various oral hypoglycemic agents, including new glucagon-like peptide-1 (GLP-1) receptor agonists like semaglutide, tirzepatide, and liraglutide. Artificial intelligence (AI) tools such as ChatGPT (OpenAI, San Francisco, United States) and Google Gemini (Google DeepMind, London, United Kingdom) provide an innovative approach to creating patient education materials, potentially enhancing the accessibility and understanding of medical information. Thus, the study aimed to compare the effectiveness of ChatGPT and Google Gemini in generating patient education brochures for semaglutide, tirzepatide, and liraglutide. Key criteria included readability, similarity, and reliability of the generated content.
The cross-sectional study design was conducted in June 2024, involving data collection from ChatGPT-3.5 and Google Gemini. Each AI tool generated educational brochures for the three medications. The responses were evaluated using Flesch-Kincaid readability scores, Quillbot similarity analysis, and a modified DISCERN instrument for reliability assessment. Statistical analysis included univariate t-tests and Pearson's coefficient of correlation via RStudio v4.3.2 (Posit, Boston, United States).
ChatGPT generated longer brochures with higher word counts compared to Google Gemini, which had better readability scores. Similarity analysis showed that Google Gemini's content had a higher percentage of overlap. Both AI tools demonstrated high reliability scores, with no significant difference between them.
Google Gemini provided more readable content, while ChatGPT produced slightly more detailed information. Both AI tools were effective in generating reliable patient education materials for GLP-1 receptor agonists. However, future research should incorporate more AI tools and updated versions for comprehensive analysis.
糖尿病管理涉及使用多种口服降糖药，包括新型胰高血糖素样肽-1（GLP-1）受体激动剂，如司美格鲁肽、替尔泊肽和利拉鲁肽。诸如ChatGPT（美国旧金山OpenAI公司）和谷歌Gemini（英国伦敦谷歌DeepMind公司）等人工智能（AI）工具为创建患者教育材料提供了一种创新方法，有可能提高医疗信息的可及性和理解度。因此，本研究旨在比较ChatGPT和谷歌Gemini在生成司美格鲁肽、替尔泊肽和利拉鲁肽患者教育手册方面的有效性。关键标准包括生成内容的可读性、相似性和可靠性。
2024年6月进行了横断面研究设计，涉及从ChatGPT-3.5和谷歌Gemini收集数据。每个AI工具为这三种药物生成教育手册。使用弗莱什-金凯德可读性分数、Quillbot相似性分析以及用于可靠性评估的改良DISCERN工具对回复进行评估。统计分析包括通过RStudio v4.3.2（美国波士顿Posit公司）进行单变量t检验和皮尔逊相关系数分析。
与谷歌Gemini相比，ChatGPT生成的手册更长，字数更多，而谷歌Gemini的可读性分数更高。相似性分析表明，谷歌Gemini的内容重叠百分比更高。两个AI工具的可靠性分数都很高，两者之间没有显著差异。
谷歌Gemini提供了更具可读性的内容，而ChatGPT生成的信息略为详细。两个AI工具在为GLP-1受体激动剂生成可靠的患者教育材料方面都很有效。然而，未来的研究应纳入更多的AI工具和更新版本进行全面分析。
