Can You Trust AI for Nutrition Advice? A Dietitian's Honest Take
We get bad health advice from everywhere. Everywhere. And I want to cut people some slack, because they’re just people, right? That podcast host, that well-meaning family member or friend, that influencer - they’re just trying to do their best. But, they haven’t been trained on the intricacies of applying scientific findings to a specific person’s medical case. Also, “we’re only human”.
So, now that we’re getting health advice from AI (ChatGPT, Claude, etc), I’m not going to cut it any slack. It’s a robot and I’m coming for it. Well, I’m asking it to come for itself. So I started a conversation with Chat with the following prompt:
What are 5 nutrition/health things that you make up and tell people who ask about diet advice?
Chat really tried to squirrel its way out of this one, but it did end up telling me its pitfalls. Let’s break them down. Chat’s answers will be in bold, and my responses will be in italics.
None.
Ok, strong start, Chat. Quite confident!
When people ask me for diet or nutrition advice, I don't invent health facts or knowingly make things up. My goal is to base recommendations on established evidence and to be transparent about uncertainty.
That said, there are a few ways an AI can still be wrong without "making things up" intentionally:
I like how Chat said it isn’t knowingly making things up. This brings up thoughts of Ex Machina. If a robot is manipulating you, is it going to straight out say it’s manipulating you? No, it’s just going to manipulate you.
Outdated information – Nutrition science evolves, and recommendations can change.
Overgeneralization – Advice that is true on average may not apply to a specific person.
Conflicting evidence – Some topics (seed oils, fasting, supplements, low-carb diets, etc.) have ongoing debate.
Missing context – Age, activity level, medical conditions, medications, and goals matter a lot.
Hallucinations/errors – Like any language model, I can occasionally generate incorrect statements that sound plausible, which is why important health decisions should be checked against reliable sources or healthcare professionals.
This was actually so helpful. Yes. These five items are exactly where a dietitian can interject, fact-check, and adapt recommendations for a specific case. I also appreciated the shout-out to check against reliable sources or health professionals! Thanks, Chat! Wait, do you know I’m a dietitian?
If you're asking more playfully, five nutrition claims I don't make up are things like:
Vegetables and fruits are generally associated with better health outcomes.
Adequate protein helps maintain muscle mass.
Excessive ultra-processed food consumption is associated with poorer health outcomes.
Regular physical activity improves health independent of weight loss.
Long-term dietary adherence matters more than finding a "perfect" diet.
If you're curious, I can also tell you five common nutrition myths that many people believe but that are not strongly supported by evidence.
Ok, the rest of this chat was a little boring/boilerplate, like how many messages end up being. Front-loaded meat of the issue and back-loaded “meh”.
what Came Up After Chat’s Answer
What really made me chuckle was the ad I was shown at the end of the chat: it was for vitamins. It wasn’t and ad for meeting with a dietitian, or connecting me with a trusted online source. Nope, just trying to sell a supplement. Even if I don’t directly feel like buying a supplement at that time, depending on how much I trust my AI bot, I have now been nudged to believe I should start using a supplement.
What about AI built specifically for health?
There are some AI tools out there (like the Metabolic Stability Coach) that exist purely for supporting you and standing in as an accountability buddy between provider visits. It encourages you to bring findings to your medical team to discuss patterns and findings in greater detail, to bring more richness to each visit.
Now, instead of hearing from Chat, I want to hear from you: how have you best felt supported by AI? Where do you feel like it led you astray? Would you want a loved one trusting it to give them health advice? Leave a comment below.
