Artificial intelligence (AI) has become an increasingly common source of nutrition information. Whether for meal ideas, calorie calculations, or general nutrition questions, many individuals are turning to tools like ChatGPT for guidance.
As AI use in this space continues to grow, we wanted to better understand its capabilities and limitations within clinical nutrition practice. To explore this, our RD team tried AI, and here’s what we’ve experienced:
1. ChatGPT does not utilize the Nutrition Care Process
When we used ChatGPT for nutrition-related questions (such as how many calories a 16 year old soccer player would need to maintain his weight), it relied on generalized data and equations rather than individualized clinical assessment. It could not account for medical history, medications, labs, lifestyle, or personal goals in the way a Registered Dietitian uses the Nutrition Care Process to guide and adjust treatment.
2. ChatGPT produces inconsistent calorie and macro calculations for recipes and meal plans
When we entered full recipes, calorie and macronutrient estimates varied based on how ingredients were categorized, how serving sizes were interpreted, and how recipes were structured. Compared to hand calculations using standardized nutrition databases, the values did not consistently match and were sometimes significantly higher or lower. In meal planning examples, ChatGPT also did not reliably convert foods into accurate exchange lists or consistent calorie totals, resulting in incorrect or mismatched planning outputs.
3. ChatGPT does not consistently apply evidence-based nutrition context
ChatGPT generated responses based on pattern recognition rather than critical evaluation of scientific literature. In our experience, this sometimes resulted in information that was outdated, drawn from mixed sources, or overly simplified. Additionally, because AI pulls information from a variety of online sources (i.e., scientific journals, other websites, personal blogs, social media, etc.), it could not always decipher trends from truth.
4. ChatGPT provides citations that are not always verifiable
When citations were included, they were sometimes incomplete, inaccurate, or difficult to trace back to original sources, making it challenging to confirm reliability. In some cases, links or references did not clearly match the information provided or could not be verified through the original source.
5. ChatGPT may reinforce disordered eating patterns
While some safeguards are in place, responses can still shift toward restriction or an increased focus on calories depending on how questions are phrased. Because ChatGPT does not assess psychological factors or eating disorder risk, it may generate guidance that is not appropriate for all individuals and could be potentially harmful in certain contexts.
Final Thoughts
AI can be a helpful resource for general nutrition education, meal ideas, and basic information; however, it operates without clinical oversight, individualized assessment, or the ability to apply professional judgment to complex health situations. It should be used with an understanding of these limitations and not as a replacement for individualized medical nutrition therapy.
Registered Dietitians apply nutrition science alongside a full understanding of a person’s health history, lifestyle, and goals to provide care that is tailored, evidence-based, and adaptable over time.
