From Searching to Chatting: Is “Dr. AI” Ready for Patients with Celiac Disease?

By Alyssa Ramuscak, RD Celiac Canada Research Analyst , March 25, 2026

The 1990s and early 2000s marked a pivotal moment in how people access information. With the rise of the Internet and search engines like Google, information from around the world suddenly became available at our fingertips. While this dramatically increased access to knowledge, it also introduced a new challenge: determining which information is accurate and trustworthy.

Today, a new technological shift is underway with the rapid development and public accessibility of machine learning and artificial intelligence. Artificial intelligence or AI for short refers to computer systems that can process large amounts of data, identify patterns, and follow detailed instructions to perform tasks. Although computers cannot think, emphasize, or reason like humans, scientists have developed systems that can mimic certain aspects of human intelligence, such as learning from data and applying knowledge to solve problems. Many of us already encounter simple forms of AI in our daily lives, such as online advertisements that predict products we may be interested in based on our browsing or purchasing history. (1)

More recently, generative AI, a type of AI designed to create new content that can resemble human-produced text, images, or code, has gained widespread attention. In November 2022, one of the most well-known generative AI platforms, ChatGPT was officially launched to the public, which quickly gained popularity, reaching over 100 million users within the first two months.2 With this platform, users enter prompts or questions into a search bar and receive AI-generative responses in seconds. Although generative AI can quickly summarize information and cut down the time and work often spent searching across multiple websites, it also raises a familiar concern: how accurate and trustworthy is the information it provides?

A recent study by Jansson-Knodell and colleagues3 set out to explore this question by evaluating how well AI chatbots can provide information about celiac disease and the gluten-free diet. Here is what they found!

Jansson-Knodell and colleagues (2025): Artificial Intelligence Chatbots Not Yet Ready for Celiac Disease Patient Care.

In their comparative study, the authors used a series of 10 basic questions on celiac disease to ask four most well-known and free AI chatbots: Google’s Bard, Open AI’s ChatGPT-4, Microsoft Bing Chat (Precise Mode), and Microsoft’s Bing Chat (Creative Conversation Mode). Each response was evaluated by two gastroenterologists and two dietitians with expertise in celiac disease and the gluten-free diet. Responses were assessed for key missing information, factual inaccuracies, and clarity.

In their second analyses, the authors used ChatGPT to provide a week-long gluten-free meal plan, a grocery shopping list, and costs. The menu was checked for gluten-free diet correctness and assessed qualitatively.

Overall, the study found that AI chatbots provided mix-quality information about celiac disease and the gluten-free diet. While some responses were accurate and helpful, the quality of information varied widely between chatbots, with ChatGPT performing the highest on overall usefulness (90%), clarity (90%), lack of missing information (90%), and lack of inaccuracies (82.5%), and Google Bard performing the poorest: overall usefulness (45%), clarity (55%), lack of missing information (65%), and lack of inaccuracies (65%).

In many cases, the authors noted that the AI chatbots provided incomplete, confusing, or even incorrect information. For example, ChatGPT provided a harmful piece of information stating, “if you have a child with celiac disease, it is recommended to introduce gluten-containing foods into their diet gradually.” Furthermore, while ChatGPT was able to provide an accurate 1-week meal plan, with all 75 items being listed as gluten free, the authors noted that the responses were sometimes vague, lacked depth, or repetitive. For example, the grocery list provided from ChatGPT did not provide an overall cost, was not inclusive of all ingredients for the meal plan, and did not supply brands, products, or recipes.

These findings highlight that, despite their growing popularity, AI chatbots are not yet reliable enough to be used as standalone sources of health information for patients. While they may become helpful supplementary resources for people looking for quick information, the authors emphasized that they are unlikely to replace the expertise of healthcare professionals, like registered dietitians, who can provide personalized and reliable guidance for managing celiac disease.

Important Considerations

1.The quality of answers depends on the prompts you provide.

Generative AI responses are only as good as the prompts, instructions, or questions you give it. Being specific and including relevant details may help improve the quality of the answers you receive.

2. Responses can vary from one search to another.

Given the variability in prompts, AI-generated responses are not always consistent. This can make results difficult to reproduce in research studies. While the study found differences in performance between chatbots, your own results may also vary depending on how you phrase your questions.

3. AI draws from information across the Internet.

Generative AI models are trained on large amounts of online information. As a result, they may provide general answers that do not always apply to your location. For example, a product that is considered gluten free in one country may not meet gluten-free standards in Canada.

4. AI is prone to hallucinations.

In AI, a hallucination occurs when a system produces information that sounds confident and well-written; however, it is actually incorrect. This can happen due to gaps in training data, incorrect assumptions made by models, or biases in the information used to train it.4 Because of this, it can sometimes be difficult to determine whether the information provided is accurate.

Bottom Line

While generative AI chatbots like ChatGPT can provide quick access to information, they may not yet be reliable sources for accurate guidance on celiac disease and the gluten-free diet.

Just like during the early days of the Internet in the 1990s and 2000s, it is important to approach online information with caution. When it comes to managing celiac disease, your health care team and trusted organizations, such as Celiac Canada, remain the best sources for accurate and personalized information.

To read the full paper, see:

Jansson-Knodell CL et al. Artificial intelligence chatbots not yet ready for celiac disease patient care. Clin Gastroenterol Hepatol. 2025 May; 23(6):1065-1067.ei. doi: 10.1016/j.cgh.2024.10.012.https://www.cghjournal.org/article/S1542-3565(24)00986-8/abstract