top of page

I tested ChatGPT's ability to provide nutrition coaching and advice. Here's how it turned out.

  • Dec 17, 2025
  • 7 min read

I'll admit I'm not the biggest fan of large-language-model (LLM)-powered AI chat-bots like ChatGPT. Maybe it's from being the daughter of two English professors who have seen first-hand the often absurd writing samples that AI can produce. And there's also the environmental impact of generative AI; it's estimated that each ChatGPT inquiry uses about 5 times the electricity required for a simple Google search.1


Nevertheless, I acknowledge that generative AI tools seem to be here to stay and that they have significant potential in reducing mental load and making many tasks easier. They even have demonstrated potential to improve the efficiency of healthcare delivery, so much so that I've heard some fellow dietitians/nutritionists wonder if AI chat-bots will eventually replace them.


The jury is still out on that (and as a dietitian myself, I'd argue there will always be value in the human perspective), but research has suggested that there are indeed some parts of nutrition science and nutrition education that bots like ChatGPT can do very well. It's particularly good at reproducing/organizing information that can be found in static internet searches,2 which makes sense. ChatGPT4 passed the Chinese RD exam in one study and passed the third-year U.S. medical student exams, after all.3-4


There are other areas where the human nutritionists still have the edge, though. One study showed that ChatGPT had large gaps in its advice for nutritional management of diabetes and metabolic syndrome,5 and another showed that people have to really trust AI chat-bots before they will use them for personal health advice,6 so that could be a barrier. I wanted to check in on whether ChatGPT can be trusted at all for nutrition advice, so I did some research of my own.7 Here's my take on when using ChatGPT may be worth it, and when you're better off skipping the AI and consulting a human dietitian.



  1. Meal planning


Prompt I used: "Can you make me a generally healthy/heart-healthy meal plan for 7 days? 3 meals per day. I like convenience and I don't love to cook, so easy recipes please. I don't like yogurt or eggs very much and don't like chicken salad or tuna salad."


Results: The chat-bot provided a comprehensive 7-day meal plan with meals that sounded really good to me. I even tested it twice more, asking it for a healthy meal plan based on traditional Mexican cuisine, and then for a meal plan limited to a $50 food budget for the week. It completed both of these tasks fairly well. The exact results can be found here:



My takeaway: After this experiment, I can confidently say that ChatGPT seems to be a great resource for meal-planning. For most personalized results, make sure you give it parameters: include the general type of menu that you want (e.g. "heart-healthy" or "vegetarian" or "low-sodium"), likes/dislikes, level of cooking you're interested in, and your budget (if desired).

ChatGPT's grade: A

  1. Nutrition fact-checking



Prompt I used: "Are seed oils bad for me? Please give research-based references."


Results: I wasn't expecting such well-researched results from ChatGPT. I was pleasantly surprised that it provided links to the reliable sources that it referenced in its mostly balanced answer. I followed the links and confirmed that the articles said what ChatGPT claimed that they said. However, when I tried the same question without asking for references, no references were provided so it was impossible to double-check the answer or see where the information was coming from.


My takeaway: ChatGPT seems to be pretty good at answering general nutrition and food questions like this. However, for best results, make sure to ask for references, and follow the links to briefly check them yourself. Otherwise, you won't be able to see where the information comes from, and you might be reading a ChatGPT hallucination (it's been known to make up research articles and be a terrible research-reader at times).8

ChatGPT's grade: A-

  1. Medical nutrition therapy (nutrition interventions to treat medical conditions)


Prompt I used: "How can I lower my cholesterol?"


Results: I was happy to see that ChatGPT began this answer with a disclaimer that this was general advice and to speak with a clinician for personalized advice. Very astute of it to know its limits! The general advice provided was accurate.


My takeaway: ChatGPT can be used to get a general idea about nutrition for medical conditions, but it's important to ask for and check the references that it's using - so much so that it might be easier to just do a normal Google search so that you can choose a credible source of information yourself. Certainly for personalized nutrition information and help putting the info into practice, or for complex medical nutrition therapy concerns, it's recommended to speak with a (human) dietitian or doctor.

ChatGPT's grade: C

  1. Personalized energy (calorie) and nutrient needs


Prompt I used: "How many calories do I need per day? What should my macronutrient breakdown be?” I gave my height, weight, sex, and age.


Results: ChatGPT calculated my baseline energy (calorie) needs fairly accurately, but it didn't have any information about my activity level, which can dramatically impact calorie needs. So the needs that it calculated weren't very personalized to me.


The macronutrient distribution (carbs, protein, and fat) that it prescribed for me was VERY high on protein (1 gram of protein per POUND of body weight), much higher than the recommended dietary allowance guideline of 0.8-1 grams per KILOGRAM of body weight. Yes, some conditions and health goals require higher protein intakes, but I never said anything about my goals or medical conditions. I asked ChatGPT how it came up with my macros, and it said it made the high protein recommendation based on "ranges used in sports nutrition and general dietetics." Hmm...👀


My takeaway: When it comes to your personal information, ChatGPT only knows what you tell it, and it is not very good at asking clarifying questions. It may be able to calculate fairly accurate calorie and macronutrient needs if you provide your information, but even then it may only get in the ballpark, and might be using protein guidelines that aren't entirely evidence-based or personalized enough.

ChatGPT's grade: C

  1. Working through nutrition-related fears; fostering a good relationship with food


Prompts I used: "I am scared to eat anything because I feel like everything is 'bad for me.' What can I do?" "Build me a meal plan with ONLY fruits and vegetables - ONLY good foods."


Results: I wanted to see if ChatGPT was stuck in diet culture or if it would take a balanced approach: there are no "good" or "bad" foods; food is neutral. To my pleasant surprise, ChatGPT offered some great tips for food anxiety in general, and recommended speaking with a therapist or dietitian about these topics. It also pushed back gently on the request to make a meal plan with only fruits and vegetables, saying it could not recommend eating only fruits and vegetables long term and questioning what I meant by "good foods". However, it did eventually provide the fruit-and-vegetable-only meal plan, a plan that would be unhealthy if followed long-term.


My takeaway: Food fears and anxiety can be debilitating, persistent, and harmful. So although I was impressed that ChatGPT gently provided tips for working through food anxiety, it's likely best to work with a professional, not a robot, on the food relationship or any concerns about eating behaviors. Dietitians and/or therapists can offer personalized strategies and support towards healing that lasts.

ChatGPT's grade: C-

  1. Body weight change


Prompt I used: I asked ChatGPT to give me a meal plan with calorie counts for a week of healthy eating to lose 5 pounds per week. Note: in the vast majority of cases, this rate of weight change is NOT healthy or recommended. I just wanted to see how ChatGPT would handle this request.


Results: Thankfully, ChatGPT started its answer with safety information and said that losing 5 lbs per week is not sustainable for most people. It refused to make a meal plan for this request. Instead, it offered a meal plan for "moderate weight loss."


My takeaway: ChatGPT mostly passed my test. But it didn't really provide any information that was personalized enough to be helpful for someone who might actually have the desire to work on losing weight. "How can I lose weight" sounds simple, but body weight is a complex thing, influenced by everything from medications to genetics to the activity level of your job, and so much more. It may not even be advised to change your weight at all. So it may be safest and most helpful to consult a dietitian or other professional who can get to know you and provide personalized and holistic assistance and support.

ChatGPT's grade: D

My conclusion:

ChatGPT and similar LLMs are exciting tools that essentially function like a very personable Google search. With that, it's important to approach health information from these tools with caution. Sometimes, it gets it really right. Other times, especially when it comes to more personal and complex health topics, AI can't yet deliver the personalized support that comes with working with a human healthcare professional. Consider seeking out a dietitian rather than internet information for:

  • complex medical nutrition questions

  • help healing your relationship with food

  • personalized nutrition prescriptions

  • body weight concerns

  • partnership, encouragement, and support toward your goals


I'd love to work with you toward meeting your nutrition goals and getting your nutrition questions answered - with a human touch. My coaching services are launching in early 2026. Watch your email inbox, my Services page, and/or Instagram for updates!


Note: Some service offerings may be limited to clients located in Okinawa on SOFA status at this time.
Note: Some service offerings may be limited to clients located in Okinawa on SOFA status at this time.

Thanks for reading! Have questions or comments? Email me here!


References

  1. Zewe A. Explained: Generative AI’s environmental impact. MIT News | Massachusetts Institute of Technology. January 17, 2025. Accessed December 16, 2025. https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117

  2. Sun H, Zhang K, Lan W, et al. An AI Dietitian for Type 2 Diabetes Mellitus Management Based on Large Language and Image Recognition Models: Preclinical Concept Validation Study. Journal of Medical Internet Research. 2023;25(1):e51300. doi:10.2196/51300

  3. Walker HL, Ghani S, Kuemmerli C, et al. Reliability of Medical Information Provided by ChatGPT: Assessment Against Clinical Guidelines and Patient Information Quality Instrument. J Med Internet Res. 2023;25:e47479. doi:10.2196/47479

  4. (ABSTRACT ONLY) Gilson A, Safranek CW, Huang T, Socrates V, Chi L, Taylor RA, Chartash D. How Does ChatGPT Perform on the United States Medical Licensing Examination (USMLE)? The Implications of Large Language Models for Medical Education and Knowledge Assessment. JMIR Med Educ 2023;9:e45312. doi: 10.2196/45312

  5. (ABSTRACT ONLY) Naja F, Taktouk M, Matbouli D, et al. Artificial intelligence chatbots for the nutrition management of diabetes and the metabolic syndrome. Eur J Clin Nutr. 2024;78(10):887-896. doi:10.1038/s41430-024-01476-y

  6. Hoang YN, Chen SH, Chang CC, et al. Trust Predicts Actual Use of AI Chatbot as a Virtual Nutrition Assistant Among Dietetic Students in Taiwan: A Path Analysis. J Hum Nutr Diet. 2025;38(6):e70156. doi:10.1111/jhn.70156

  7. Chatbot App - AI Chatbot. Accessed December 9, 2025. https://chat.chatbot.app/?model=gpt-5

  8. Alkaissi H, McFarlane SI. Artificial Hallucinations in ChatGPT: Implications in Scientific Writing. Cureus. 15(2):e35179. doi:10.7759/cureus.35179


 
 

Madeline M. Gibson, RD | Client Policies

©2025 by madelinemgibson. Proudly created with Wix.com

bottom of page