AI Chatbots: Can I Use Them for Nutrition Advice?

By KRNC Staff

Since the introduction of new Artificial Intelligence (AI) technologies, it’s never been easier to get an answer to our most burning questions, especially those regarding nutrition and health advice. It’s common to ask AI questions about what foods to eat, how many calories to intake, how to manage a health condition or what to make for dinner tonight. Although AI can weigh in on lab results and build a meal plan, is it safe to forego professional care and trust an AI bot to manage your health? When our health is on the line, it may be a bigger risk than we think.

How Does AI Work?

AI is a technology that allows computers to make decisions similar to how humans make decisions. It works by collecting data, processing that data and detecting patterns, then generating content. AI models can be trained with certain parameters to help improve their predictions and give more helpful information.

Much of this AI technology is implemented in chatbots that use natural language processing to give human-like interactions. This can be seen on platforms such as ChatGPT, Claude, and Gemini. These tools learn from a massive database to make their predictions and give the user an answer. These databases have access to most publicly available information found online.

AI Use for Nutrition Information

With AI’s widespread use, it’s no surprise that it has entered the nutrition world. The public is primarily using AI technology for their nutrition care in two ways, chatbots and food logging apps. Chatbots are being asked by users to create meal plans, calculate macronutrients and calorie needs, and obtain personalized nutrition recommendations. Food logging apps, both preexisting and new, have implemented AI technologies into their platforms. Using AI technology not only gives these apps the capability to integrate chatbots, but also tools such as calorie need calculators, dietary intake assessments based on photos, and meal ideas.

Potential Benefits of Using AI Technology for Nutrition Care

One of the perceived benefits of AI use for nutrition recommendations is accessibility. Users believe it decreases the barriers to navigating nutrition care, such as time and cost. AI tools skip over in-depth assessments of medical history and nutrition needs and give immediate feedback on diet plans and nutrition changes. AI for nutrition care has been shown to help in short-term settings but fails to help users make long-term changes when used alone. Unfortunately, the benefits of AI for nutrition care seem to end here due to its limitations.

Where AI Gets It Wrong

Because of the nature of AI tools, the nutrition advice presented is at risk of being inconsistent and unreliable. This has been shown in AI bots giving misinformation, AI technology miscalculating nutrients, and limited interpersonal recommendations.

Studies researching nutrition recommendations generated by ChatGPT have found several instances where incorrect or incomplete information was given. A 2023 study by Niszczota and Rybicka found that ChatGPT confidently generated menus and included almond milk when a nut allergy was specified. Mistakes were made in 4 out of the 56 prompts. Another 2023 study by Chetelan et al. showed that ChatGPT generated flawed medically tailored meals when promoted. The study asked ChatGPT to create a meal plan for a patient with Type 2 Diabetes. This meal plan was in accordance with dietary guidelines. However, when asked to create a meal plan for a patient on hemodialysis, it suggested foods high in phosphorous and potassium, which goes against the medical nutrition therapy (MNT) recommendations for hemodialysis patients. Both studies showed that ChatGPT typically supplies accurate general nutrition information, but it is not perfect and still frequently includes mistakes which should not be taken lightly.

In addition to these mistakes, AI tools have also been found to frequently miscalculate calories. Most frequently, apps that use photos to generate information on daily food intake often get it wrong. The AI tools are often unable to distinguish visually similar foods, such as tofu and paneer cheese or ice cream and Greek yogurt. These tools are also unable to accurately determine the serving sizes shown in uploaded photos. This leads to the AI tools miscalculating the grams of each macronutrient: fat, protein, and carbohydrates. When these are miscalculated, the total calorie counts are incorrect. Because of these inaccuracies, apps give misguided direction to its users. Inadequate or inaccurate macronutrient calculations can have serious health consequences, especially when this information is delivered with confidence and no warning to verify it with a professional.

Lastly, AI tools are not humans. They don’t have the same capabilities to create personal connections and provide the support needed for sustainable behavior change. They lack the ability to create accountability, understand cultural and social dynamics of foods, make safe judgement calls, and educate with evidence-based care. While studies show that users can make improvements to their diet in a short-term period, AI tools alone are not effective for making long-term changes. Without the decision-making skills of a human professional, AI tools have also been found to make recommendations that support disordered eating habits. AI food logs tend to encourage extreme calorie deficits and reward restriction, making them dangerous and unreliable. Because AI lacks interpersonal skills and empathy, it is unable to effectively give safe and sustainable nutrition care.

It should also be noted that there are no safeguards in place to protect your information in AI platforms. This means that the health information that you input could be leaked, stolen, or sold without your consent. Unlike medical providers, platforms such as ChatGPT do not need to follow HIPAA laws. This lower protection should increase wariness around the use of AI for health information.

Although AI tools have many downfalls, they do not need to be written off altogether. This ever-growing and changing technology can be used safely in conjunction with professional healthcare.

How to Safely Use AI for Nutrition Advice

Nutrition recommendations from AI tools should never be your sole dietary guidance. A Registered Dietitian Nutritionist (RDN) is the most qualified professional to help you make informed and personalized nutrition choices. Meeting with an RDN can help you manage nutrition for certain medical conditions, heal your relationship with food, or improve general healthy eating habits. RDNs are uniquely qualified to not only share nutrition education, but to also help you navigate barriers to applying nutrition recommendations. Behavior change requires more than just knowing what to do, as suggested by a chat bot.

So where can AI be used in nutrition care? Think small and collaborative uses. Research shows that AI can be an effective nutrition care tool when used in conjunction with nutrition therapy from an RDN. AI technologies should not be creating large scale plans of action such as macronutrient calculations, calorie counts, or comprehensive Medical Nutrition Therapy. Instead, AI tools can help implement these personalized care plans formulated by an RDN.

For example, let’s say part of your RDN’s care plan for you is to include 25g of daily fiber. Maybe you talked with your RDN about adding chia seeds to your daily yogurt to help with this, but you ran out of chia seeds. You could use an AI chat bot to help identify a suitable substitute that you have at home by entering a prompt such as, “I’m trying to add more fiber to my diet. I have the following ingredients: rice cereal, flaxseeds, chocolate chips, watermelon, dried blueberries, and cinnamon. Which ingredients should I add to my yogurt to get the most fiber?”. The AI bot will likely tell you to add the dried blueberries and flaxseeds because it has that information in its database. You can then discuss this swap in more detail with your RDN at your next visit.

More examples of helpful AI use include creating grocery lists, coming up with meal ideas, and navigating logistics around food. That could look like inputting into a chatbot a foods list recommended to you by your RDN alongside a list of foods that you already have at home. Based on this input, you could ask follow-up prompts that request recipes to be made using these criteria, then make a grocery list from the recipes you select, and even ask what grocery stores in your area have the best deals on needed ingredients. AI can even come into play when you’re wondering which recipe to prepare first to prevent spoilage, the proper cooking methods for food, or best way to store leftovers.

In these examples, the AI technology was not driving the course of nutrition care but rather giving supplemental information and support between sessions with a dietitian. This supportive and collaborative approach yields safer and more effective nutrition outcomes than the use of AI alone. This collaborative approach should still be used with high caution. AI tools often fail to cite where their information came from, which can give misleading information.

The Bottom Line

We live in a world with an abundance of information at our fingertips, now more than ever! It’s easy to become over-reliant on AI for many parts of our lives, including our health. AI technology can be a helpful tool for nutrition in the right circumstances, but never alone. With the popularization of virtual RDN visits and an increase in insurance coverage for nutrition care, many have growing access to safe and reliable care. Ultimately, RDNs can give you the tools you need to make empowered nutrition choices for your health rather than taking risks with AI tools.

More Information

For additional resources for healthy eating, check out our upcoming Cooking Classes! Find delicious and healthy recipes on our Recipes page! More health tips are also available at the College of Health and Human Sciences Pinterest board. Lastly, don’t forget to sign up for the KRNC monthly newsletter!