Artificial intelligence (AI) is changing the way we access health information, but how reliable is it for people living with celiac disease? A team of researchers from Italy, the United States, the United Kingdom, and Germany explored this question by testing three leading AI tools (Open AI’s ChatGPT‑4, Anthropic’s Claude 3.7, and Google’s Gemini 2.0) on their ability to provide information about celiac disease. The researchers examined accuracy, clarity, and ease of understanding across topics such as diagnosis, symptoms, diet and nutrition, and daily disease management, raising important questions about whether these tools are ready for patient use.
In this study, four gastroenterologists and two dietitians reviewed answers generated by each model. The study found that while all AI models were capable of generating medically accurate information, they differed in important ways that affect patient education. Among the three, Google’s Gemini 2.0 stood out with higher scores for both accuracy and clarity. All three models exceeded the recommended sixth- to eighth-grade reading level for health education, which matters because overly complex information can lead to misunderstandings about managing a strict gluten-free diet.
The researchers also highlighted that misinformation remains a challenge in across all three AI models, with 13-24% of responses in this study containing errors. One area of concern across all three models was practical dietary management, particularly cross-contact prevention, where AI performance was lower. Diet education in celiac disease is complex and requires combining food science, culinary practices, and clinical guidelines, while also accounting for individual household setups and current labeling regulations. These findings suggest that AI developers should prioritize specialized training data and dietitian-reviewed protocols, and that healthcare providers should review AI-generated advice carefully before patients rely on it.
Despite these limitations, AI has potential to reduce the burden on healthcare providers and improve patient access to information. Studies show that many patients are uncertain about various aspect of the gluten-free diet, causing them to turn to online sources for guidance.
“So many of our patients use AI for quick answers for many of their health questions,” shared Anne Lee, EdE, RDN, LD, study author and member of the Celiac Disease Foundation Dietitian Advisory Council (CDF-DAC). “While the three models are quick and accessible tools, it is worrisome that almost 25% of the information contained errors or out dated information. This was especially true in the area of the GFD. There were many errors on the potential for chance exposure, cross contact, and foods allowed on the diet. This study is so important for clinicians so we may be better prepared to educate and direct where our patients obtain health information.”
Rather than discouraging AI use, this research points to the value of structured integration with clinical oversight. Thoughtful model selection and provider review could allow AI-generated materials to serve as a starting point for celiac disease education and improve access to accurate information.
As technology continues to evolve, AI has the potential to complement patient education and support better disease management. Today, the Celiac Disease Foundation is already leveraging innovative tools alongside expert-developed resources to help individuals thrive with celiac disease. Programs like Foundations of Celiac Disease provide a free virtual education series for those newly diagnosed or without access to specialized care. In addition, the Foundation’s Innovations in Culinary Medicine program offer quarterly sessions led by field experts, covering everything from cutting-edge research to practical nutrition strategies for managing gluten-related disorders.
Learn more and register for Foundations of Celiac Disease here and Innovations In Culinary Medicine here.