A case research reveals the actual risks of utilizing AI for diet recommendation.
By now, you are possible conscious of Synthetic Intelligence, together with chatbots comparable to ChatGPT. These functions have made their approach into practically each nook of our lives, promising to be your individual private assistant. Naturally, this has led many to query whether or not ChatGPT and AI can be utilized to assist with diet.
Or, are you able to truly belief AI to jot down a food regimen plan? A current case report in Annals of Inner Drugs: Scientific Instances suggests the reply ought to include a heavy warning label. Because of a massive misunderstanding of how AI Chatbots work, blindly getting any kind of well being and food regimen recommendation can truly land you in bother, together with the hospital.
Can you utilize ChatGPT for diet and coaching recommendation?
5 Key Factors You Want To Know:
|
How ChatGPT Gave Eating regimen Recommendation That Led To The Hospital
In 2025, a broadcast case research in Annals of Inner Drugs: Scientific Instances paperwork how a 60-year-old man developed bromism, a harmful toxidrome attributable to extra bromide ingestion (Eichenberger et al., 2025).
What was as soon as frequent within the twentieth century, bromism may end up in varied well being points, together with:
- Set off hallucinations,
- Paranoia
- Fatigue
- Profound electrolyte imbalances.
So, how did this man find yourself poisoned? And why would he eat bromide?
Effectively, he did it to himself as a result of ChatGPT advised him to.
How Did ChatGPT Poison a 60-12 months-Outdated?
After being admitted to the hospital, the person confessed to having varied beliefs on diet, together with distilling his personal water at dwelling and varied food regimen restrictions. He appeared to get caught up in the entire “X ingredient is dangerous for you” arguments we see on-line.
Upon studying the “risks” of sodium chloride (i.e., desk salt), he consulted ChatGPT for recommendation on lowering it from his food regimen. Sadly, ChatGPT advised him he may change sodium chloride with sodium bromide.
Since ChatGPT is portrayed as being an infallible piece of expertise, he listened. He went on-line, purchased some sodium bromide, and started changing his salt with it.
Over three months, this “AI-guided” experiment led to extreme psychiatric and metabolic problems, touchdown him within the hospital. His bromide ranges have been discovered to be 200 occasions above regular. Fortunately, after fluids and remedy, his signs resolved.
Why Would ChatGPT Advocate Bromide? Understanding How AI Chatbots Really Work.
- Chatbots function as an LLM that works by studying patterns and predicting phrases
- AI chatbots are susceptible to hallucinations once they present incorrect particulars and even invent false info
- A lot of hallucinations have been described in literature.
One of many main points that causes confusion when utilizing chatbots like ChatGPT is a misunderstanding of how they work. This is largely due to how they’re introduced to the general public.
Basically, most individuals assume these AI chatbots work like a large laptop, analyzing all of the out there info and formulating one of the best reply.
This is not the case. Removed from it.
AI chatbots (ChatGPT, Claude, Gemini) function as a Massive Language Mannequin (LLM) that predicts phrases in a sequence, based mostly on patterns from an unlimited quantity of textual content. To do that, it is first fed a ton of information from articles and books to show it issues like grammar, details, reasoning constructions, and writing kinds.
It may possibly then use all this info to reply questions you ask it. Nevertheless, right here lies the problem. LLMs do not actually “perceive” info; it is simply actually good at predicting what phrases seem collectively.
They do not have reasoning expertise in the way in which we expect, particularly with new info. This can also be why you hear of “hallucinations” when ChatGPT makes up info. Hallucinations are an actual phenomena that attain farther than Reddit boards and are documented in scientific literature (Ahmad et. al, 2023).
And so they occur lots; way more than some appear to wish to consider. A big research from Chelli et. al (2024) discovered that varied chatbots hallucinated 28.6% – 91.4% of the time when citing scientific research. This ranged from getting authors mistaken to outright inventing research.
ChatGPT is not “mendacity”, it is simply not designed to supply info it would not know. Worse, it would hardly ever say “I do not know”. Because it predicts letters, it does its job, and no matter comes out, comes out.
On this case, the person possible merely requested about changing sodium chloride, and ChatGPT supplied a solution based mostly on info with cleansing provides.
The Actual Risks of Counting on AI for Eating regimen Recommendation
- Ideally, a consumer has fundamental data of the knowledge they’re looking for
- In its present state, a consumer should fact-check AI
- Merely understanding that AI chatbots like ChatGPT make errors is essential in optimizing their use.
We’re not attempting to attenuate this expertise—it is extremely helpful in the fitting circumstances, and it’ll nearly definitely enhance over time. However opening it as much as most people with false expectations poses actual risks. This case makes that clear.
- AI lacks medical judgment. A human nutritionist or doctor would by no means suggest bromide as a salt substitute. AI would not distinguish between secure and unsafe functions—it simply generates textual content that “match“ (Walsh et. al, 2024)
- AI requires knowledgeable enter. Maybe the affected person did not ask particularly for a dietary substitute, which highlights the problem: customers should know the way to phrase the fitting query. With out that baseline data, the output will be dangerously deceptive.
- AI decontextualizes info. An announcement legitimate in a single setting (chemistry, manufacturing) could also be lethal in one other (food regimen and well being). Water can put out a fireplace—however pour it on a grease hearth and you may make issues worse.
- Sufferers and weak teams are in danger. These experimenting with restrictive diets, fast fixes, or those that lack technological literacy could take AI recommendation actually with out understanding the dangers.
- AI has a bias to please. ChatGPT and comparable fashions are tuned to give solutions customers will settle for. That may result in cherry-picked, one-sided replies: a vegan would possibly hear that plant-based consuming is the optimum way of life, whereas a keto fanatic is likely to be advised keto is greatest for muscle retention. The mannequin adjusts to the consumer’s framing, not goal medical fact.
It is essential to grasp these limitations when utilizing it to make selections in your life (Walsh et. al, 2024)
Ought to You Use AI for Eating regimen Recommendation?
- A consumer ought to be conversant in the subject in an effort to determine false info.
- All the time fact-check the knowledge. All the time.
- At this level, AI doesn’t appear to be a viable different for health and diet recommendation, particularly for these new to health and weight-reduction plan.
Take into account that an odd attribute of ChatGPT, and comparable chatbots, is that completely different folks can report very completely different experiences. Some report it is spot-on with solutions, whereas others have claimed it is turn into unusable as a consequence of its solutions.
Paradoxically, that is just like human trainers and nutritionists. The distinction is that individuals usually know to be cautious of the recommendation they get from different people. Consequently, they may depend on evaluations, have a look at completely different sources, or at the least use some well being skepticism.
Nevertheless, the key downside with AI and chatbots giving health and diet recommendation is that individuals have mistakenly been led to consider they’re flawless. Folks consider they’re hyper-complex processors that present solutions with 100% accuracy. Sadly, they do not.
Actually, many researchers have mainly said that utilizing AI chatbots and ChatGPT is ineffective. In an article printed in Schizophrenia, Emsley (2023) warns;
“…use ChatGPT at your individual peril…I don’t suggest ChatGPT as an help to scientific writing. …It appears to me {that a} extra speedy menace is (it is) infiltration into the scientific literature of lots of fictitious materials.”
This does not imply this expertise is junk (some could say that, although) or ineffective. It simply means a consumer should have the fitting expectations when utilizing it. Extra importantly, this requires the consumer to have some fundamental data of what they’re asking about.
How will you know if one thing sounds mistaken if you do not know what ought to sound correct?
AI instruments may also help summarize diet ideas, generate meal concepts, and clarify fundamental dietary pointers. However they need to by no means change skilled medical recommendation. With out guardrails, AI can produce strategies that sound authoritative but are incomplete, deceptive, and even dangerous.
Remaining Classes On AI Chatbots, Vitamin, and Health
Sure, AI can technically “write“ a food regimen plan—however ought to it? Not with out oversight. The bromism case is a sobering reminder that whereas AI is highly effective, it is not a health care provider, dietitian, or well being coach. As these instruments unfold, the actual duty falls on each builders and customers to strategy AI well being recommendation with warning, skepticism, and significant assessment.
And that is the crux of the problem: we will not actually “blame“ ChatGPT itself. The higher accountability lies with the builders, influencers, and media who oversell this expertise as greater than it’s, at the least for proper now.
| What You Want To Do: All the time fact-check well being recommendation and seek the advice of a professional skilled earlier than making dietary modifications. AI could be a software, however it ought to by no means be your solely information in the case of your well being. |
And at all times fact-check.
Reference
1. Audrey Eichenberger, Stephen Thielke, Adam Van Buskirk. A Case of Bromism Influenced by Use of Artificial Intelligence.AIM Scientific Instances.2025;4:e241260. [Epub 5 August 2025].doi:10.7326/aimcc.2024.1260
2. Ahmad, Z., Kaiser, W., & Rahim, S. (2023). Hallucinations in ChatGPT: An unreliable software for studying. Rupkatha Journal on Interdisciplinary Research in Humanities, 15(4), 12. https://www.researchgate.net/publication/376844047_Hallucinations_in_ChatGPT_An_Unreliable_Tool_for_Learning
3. Chelli M, Descamps J, Lavoué V,Trojani C, Azar M, Deckert M,Raynier JL, Clowez G, Boileau P,Ruetsch-Chelli C Hallucination Charges and Reference Accuracy of ChatGPT and Bard for Systematic Evaluations: Comparative Evaluation J Med Web Res 2024;26:e53164 https://www.jmir.org/2024/1/e53164
4. Emsley, R. ChatGPT: these should not hallucinations – they’re fabrications and falsifications. Schizophr 9, 52 (2023). https://doi.org/10.1038/s41537-023-00379-4
5. Walsh DS. Invited Commentary on ChatGPT: What Each Pediatric Surgeon Ought to Know About Its Potential Makes use of and Pitfalls. J Pediatr Surg. 2024;59(5):948-949. doi:10.1016/j.jpedsurg.2024.01.013
