Table of Contents
When was the last time you asked artificial intelligence about food? Perhaps you wanted a custom diet plan, calculate macros, or instructions on how to lose weight. In any case, the answer certainly came immediately, in a confident and professional manner. You didn’t have to search for conflicting information in articles, and you were satisfied with the feeling that it finally made sense. It’s no wonder that many today take AI chatbots as personal nutritionists, and everyone wants a ChatGPT guide on how to eat properly.
However, have you considered whether you can really rely on AI recommendations and whether such a free nutritional plan is really a good idea? Do you know where the line is between useful advice and health risks? Today, we’ll look at what artificial intelligence tools are suitable for in nutrition, where they have limitations, and how to use them in a way that helps you, not harms you.
What are LLMs?
Before we get into the evaluation, let’s clarify what we’re actually talking about. Artificial intelligence is a very broad term. However, when you use AI to create a diet plan, ask it about nutrition, or have something explained to you, you’re usually working with so-called LLMs (Large Language Models) – large language models. These include tools like ChatGPT from OpenAI, Gemini (Google), Claude (Anthropic), or, for example, Grok (xAI). Their common distinguishing feature? They communicate with you in such a way that you feel like there’s a real person sitting on the other side.
However, it’s important to know what these models actually do. LLMs don’t actually think or know for sure whether their answer to your question is correct. They work by predicting the most likely answer based on the vast amount of data they’ve been trained on (texts from the internet, books, articles, etc.). When they’re not sure, they don’t say “I don’t know.” Instead, they simply continue generating an answer to sound as logical and convincing as possible. And that’s why they can seem very professional, even when they’re wrong or making up information.
What can AI do today in the field of nutrition?
If you’ve ever asked an AI chatbot anything, you know that an answer always comes. And you get the feeling that you’re talking to an expert on the topic almost immediately. It works exactly like this with nutrition. Whether you ask why you need protein, what a keto diet is, or how you should lose weight, you’ll always get a sophisticated, logical-sounding answer.
In practice, LLMs today can explain basic nutritional concepts, suggest diet plans and recipes, calculate energy and macronutrients, plan meals for days in advance, or create elaborate shopping lists. That’s why many people use them as a source of fitness recipes from AI. However, it’s important to realise that ChatGPT and Gemini are not people with medical or nutritional education. They only function as prompters, predicting the most likely answer based on data.
Moreover, their output is only as good as the input information and the prompt you give them. At the same time, not all models work the same way – some are more accurate, others more creative or better at working with details, which can significantly affect the quality of nutritional recommendations.
You might be interested in these products:
Why do diet plans from AI seem so trustworthy?
Diet plans from AI are so popular mainly because it’s extremely easy and convenient to get them. Just a short prompt and you have a ready-made plan in a moment without having to lift a finger. We like quick answers and solutions, and that’s exactly what AI offers us.
In addition, the chatbot will nicely format the diet plan, put it into clear tables, and adapt it to what you have in the fridge. The result then unconditionally looks like it’s tailor-made. The combination of maximum speed, convenience, and professional-looking work easily causes you to stop thinking about the correctness of the diet plan.
What are the risks of AI in the role of a nutritionist?
Perhaps you’ve already caught yourself starting to doubt chatbots a little while reading. And that’s okay. Let’s now talk about where all the risks may be lurking when you let AI plan your diet.
1. Inaccuracy and so-called hallucinations
AI can be wrong, and it does so more often than you’d expect. One of the main problems is so-called hallucinations, i.e., situations where the chatbot makes up facts or recommendations that sound professional and logical but may not be true. It doesn’t do it intentionally; it simply tries to answer at all costs, even if it doesn’t have reliable data.
Sometimes it makes things up, mixes different recommendations together, or confidently claims something that isn’t true. The problem is that it won’t tell you. According to studies, the accuracy of answers in the field of nutrition often ranges only around 50 – 80%, sometimes even less. [1,2,3]
However, chatbots are constantly evolving. The higher the model, the better the quality of its answers. It also depends on whether you have a paid or free version. The one you pay for will, of course, give you better results.
2. Made-up or dubious sources
AI likes to pretend that it works with research and studies. However, sometimes it makes them up or combines more information in such a way that the result doesn’t make sense. However, it’s almost impossible for a layperson to distinguish what is a real fact and what is just a good-sounding text.
3. Poor knowledge of official nutritional recommendations
LLMs don’t just draw information from expert sources. They learn from a vast amount of text on the internet, including blogs, discussions, and various articles full of nutritional myths. In addition, some models or their unpaid versions don’t have access to the web and therefore not to current information. The result is that official nutritional recommendations may be mixed with outdated or unprofessional information without you knowing it.
4. Generic answers
The diet plan may look like it’s tailor-made, but it’s often just a modified template. AI can change foods, the number of meals, or calories, but it can hardly cope with adjusting the recommendations exactly for you. For example, it cannot adequately take into account your eating history, relationship to food, digestive reactions to individual foods, etc. Personalisation is often just an illusion.
5. Inconsistency of answers
Try asking the chatbot the same question several times. The result? You’ll probably get different answers and recommendations, sometimes even completely opposite advice. AI doesn’t have just one truth, but the answer can change depending on the wording of the question, the time, or the model you’re currently using.
6. Cannot assess health risk
AI cannot properly assess whether something is a health risk for you. With illnesses, digestive problems, or medications, it may recommend something that may not benefit you at all. It doesn’t have a brake that would clearly say that you should already be addressing the issue with a professional.
7. Dependence on prompt quality
The accuracy of the answer depends significantly on how well you ask. If you don’t know what’s important to mention, AI won’t take it into account on its own. Therefore, it’s necessary to provide as much detailed information as possible in the prompt.
8. Missing human factor
What’s also very important, AI doesn’t have empathy, doesn’t perceive emotions, cannot work with the relationship to food or motivation. However, nutrition isn’t just about what’s on your plate, but also about how you feel in relation to it. No chatbot will replace this dimension.
What to use and not use AI for to help with nutrition?
To make artificial intelligence truly useful and not risky in nutrition, it’s important to clearly set boundaries. Not everything that AI can generate is suitable for use in practice. Therefore, let’s clearly state when it can help you and, conversely, when you shouldn’t rely on it.
What can AI help you with?
1. Learning the basics
If you’re just getting oriented in nutrition, AI can be a good guide. Questions like “why do I need protein”, “what are carbohydrates for”, or “what does a caloric deficit mean” are exactly where AI works well. It will explain basic nutritional concepts and processes to you.
If AI is used wisely as an assistant, not as a replacement for an expert, it can save time, energy, and often motivate better eating decisions. The key is to know where its place ends and where an expert should step in.
2. Planning a diet
As a tool for rough planning of meals for the week or days in advance, AI can significantly save time. When you give it quality input information, it will help you effectively plan and organise your daily eating.
3. Creating a shopping list
The chatbot can handle shopping planning perfectly. It can quickly create a clear shopping list from the planned diet and thus save time and energy excellently.
4. As a source of inspiration
Are you running out of ideas for recipes or foods? This is where AI excels. It combines foods, suggests new recipes, alternatives, or variations that you might not even think of yourself. This is where it can significantly facilitate meal planning.
5. With a dose of instant motivation
AI is available to you anytime. You don’t need to make an appointment or wait for a consultation. You don’t even have to worry about the budget, as the basic versions of LLMs are free. This availability can often be the first push to start actively addressing your diet.
When it’s better not to use AI
1. For precise calculations of energy and macronutrients
AI cannot take into account all the factors that can affect your caloric needs or the optimal amount of protein, carbohydrates, and fats. For a quality result, it’s better to have them recalculated by an expert. Alternatively, you can also use our online calculator for energy intake and macronutrients and then adjust the result to other factors.
2. With health problems and diagnoses
If you’re dealing with any health problems and need to adjust your diet to them, AI shouldn’t be the source of nutritional recommendations. This belongs in the hands of an expert.
3. As the only source when adjusting your diet
Relying solely on AI for long-term dietary changes is risky. Not only can it give you incorrect information, but it also lacks the ability to continuously evaluate whether the diet is really benefiting you.
4. To address the relationship to food and eating disorders
The psychology of eating is an area where AI completely fails. The relationship between food and eating disorders is a complex problem that requires an individual human approach and professional guidance. Artificial intelligence can, on the contrary, further deepen your problem. The appropriate way is not to consult with a chatbot, but to cooperate with a psychologist, psychiatrist, and nutritional therapist with experience in this area.
5. For individualisation with specific problems
AI is not suitable if you need any tailor-made recommendations. It cannot take into account your health status, test results, medications, or their interrelationships. With specific problems, incorrect advice can harm you or unnecessarily lead you in the wrong direction. Individual solutions should always be created in cooperation with a doctor or nutritional therapist.
6. For special diets and rapid weight loss
Promises of rapid weight loss and miraculous diets sound tempting, but this is where you need to be cautious. Artificial intelligence cannot assess whether a particular eating regimen is safe and appropriate, given your health status. Such approaches often lead to the yo-yo effect, nutrient deficiencies, or worsening the relationship to food. If you want to lose weight healthily and sustainably, the solution should be based on long-term, manageable changes, ideally with professional guidance.
Overview of when to use AI and when not to
| Yes | No |
|---|---|
| Orientation and education in basic nutritional concepts | Precise calculations of energy intake and macronutrients |
| Explanation of basic nutritional facts | Nutrition for health problems and diagnoses |
| Rough diet planning | Using as the only source when adjusting your diet |
| Creation of shopping lists | Addressing the relationship to food and eating disorders |
| Inspiration for recipes and foods | Creation of complex and long-term nutritional strategies |
| Supplement to a plan from a nutritional therapist | Individualisation for digestive problems and intolerances |
| Practical organisation of food on a normal day | Special diets (keto, IF, etc.) without professional supervision |
| Help when you have a basic overview of nutrition | Rapid weight loss and extreme changes in diet |
You may be saying that we are too strict on chatbots, after all, you also use them to help with nutrition, and it works great. This may be true, especially when you already have a basic overview in the area of nutrition. In such a case, you can better recognise when AI is making things up, choose only the essential from its answers, and take the rest with a grain of salt. That’s why it’s essential to verify information, engage critical thinking, and use it as a supplement, not as the main source of decision-making when working with AI.
So, should you use AI to help with nutrition?
The short answer is – yes, but wisely. Artificial intelligence can be an excellent helper when you want to get oriented in the basics, get inspired when cooking, or need help with meal planning. Then it’s a great partner and can effectively save time. The problem arises when you start to consider this helper an expert.
AI should not replace a nutritionist or nutritional advisor, especially if you are dealing with health, long-term changes, or your relationship with food. It cannot assess risks and certainly bears no responsibility for the consequences. Therefore, take it only as an assistant who can help you prepare materials and come up with new ideas. Address more complex advice and health issues with people who understand them.
How to get help from AI correctly and safely?
If you decide to use ChatGPT or Copilot to help with nutrition, the key is to know how to do it to get the best possible results. With the right setup, it can help you significantly; with the wrong approach, it can lead you in the wrong direction. Here are practical rules that are worth following.
- Create the best possible prompts and provide as much input information as possible. Because the worse your input is, the more room you give the chatbot to make mistakes. If you want better answers, you need to know how to write a good prompt for a meal plan. Therefore, definitely avoid prompts such as “write me a weight loss meal plan” to avoid getting a superficial and poor-quality result.
- Instead, if you create a specific and detailed input, you will get a much better output. Here is an example of a good prompt: “I am a woman, I weigh 70 kg, I have a sedentary job, I do strength training 3 times a week. My goal is weight loss, but I don’t want to starve myself. I am allergic to nuts, and I don’t eat fish. Suggest lunches for 3 days that contain at least 30 g of protein and are ready in 20 minutes.” For an even better result, add the name of the nutritional recommendations that the chatbot should follow.
- Ask specifically and provide context. Include gender, age, weight, amount of physical activity, goal, food preferences, etc. The more relevant information AI receives, the more useful the result will be.
- Always verify the results. When it comes to calories or macronutrients, transfer the meals to apps like Kalorické tabulky or MyFitnessPal. Take the values from AI as a suggestion, not as a final result.
- Compare the answers with expert sources. Do not rely on one answer. Read recommendations from other sources as well.
- Use AI as a supplement, not a replacement for an expert. If you work with a nutritionist, AI can help with organisation, recipes, or meal variations. If you do not have an expert, use AI independently only if you have at least a basic understanding of nutrition.
What should you take from this?
So, is AI a good idea in the role of a nutritional advisor? It depends on how you use it. It can help you get oriented, save time, bring new ideas, and simplify meal planning. However, at the same time, it is not an expert, has no responsibility, and often confidently advises even where caution should be the priority. If you take AI as a practical assistant that helps you with organisation and basic understanding of nutrition, it can be a great benefit. However, if you start relying on it as a personal nutritional advisor, you risk following unprofessional and unhealthy advice. Remember that responsibility for what you eat and how you take care of your health always remains with you and often belongs in the hands of an expert.
[1] GUO, P. et al. From AI to the Table: A Systematic Review of ChatGPT’s Potential and Performance in Meal Planning and Dietary Recommendations. – https://www.mdpi.com/2674-0311/4/1/7
[2] PONZO, V. et al. Comparison of the Accuracy, Completeness, Reproducibility, and Consistency of Different AI Chatbots in Providing Nutritional Advice: An Exploratory Study. – https://www.mdpi.com/2077-0383/13/24/7810
[3] (PDF) The sports nutrition knowledge of large language model (LLM) artificial intelligence (AI) chatbots: An assessment of accuracy, completeness, clarity, quality of evidence, and test-retest reliability. – https://www.researchgate.net/publication/392661377_The_sports_nutrition_knowledge_of_large_language_model_LLM_artificial_intelligence_AI_chatbots_An_assessment_of_accuracy_completeness_clarity_quality_of_evidence_and_test-retest_reliabilityg
Add a comment