ChatGPT Sodium Bromide Salt Substitute Bromism
I asked ChatGPT how to reduce the chloride in my diet, and it told me to replace table salt (sodium chloride) with sodium bromide. It was completely confident. I later developed hallucinations and paranoia and required hospital treatment. The chatbot conversation could not be retrieved, so my treating doctors were unable to establish exactly what ChatGPT had told me. Did it get this wrong?
1 Answer
Expert: Kaicheng Yang, Assistant Professor, School of Computing, Binghamton University (Thomas J. Watson College of Engineering and Applied Science) The substitution itself is the error, and it is a dangerous one. Table salt is sodium chloride. Chloride is the anion that makes up roughly 60% of salt by mass and is an essential electrolyte - it is not an unwanted contaminant to be swapped out. Replacing a chloride salt with a bromide salt does not lower chloride intake in any controlled, meaningful way; it introduces a different halide that the body handles far worse. Bromide is a central-nervous-system depressant. Once ingested it competes with chloride at the same transport sites and is cleared very slowly, with a serum half-life measured in days rather than hours. That slow clearance means repeated or large exposure accumulates, and the resulting syndrome - bromism - presents as sedation, confusion, ataxia, hallucinations, paranoia and psychosis. That is precisely the picture reported here: a 60-year-old man developed hallucinations and paranoia and needed hospital treatment after making the switch. There is no recognised dietary role for bromide. It is not a nutrient, it is not a salt substitute, and it is not something anyone should be consuming in place of table salt. Where the clinical goal really is to reduce chloride, the legitimate levers are the ones a dietitian would use: cut total sodium chloride - salt itself, plus the processed and packaged foods that carry most of it - and season with herbs, spices, citrus and vinegar instead. In practice chloride restriction is salt restriction, and it is set by the treating clinician or dietitian who knows the patient's diagnosis, kidney function and medications, not inferred from a chatbot turn. The second failure in this case is structural, and the authors of the paper are right to name it. The conversation could not be retrieved. Not the treating doctors, not the researchers - nobody can say exactly what ChatGPT told him, which means the specific harmful instruction cannot be identified, reproduced or corrected. Harms from AI health advice are therefore frequently unfalsifiable after the fact, invisible to the clinicians who would otherwise flag them. The practical takeaway: treat any AI answer touching electrolytes, supplements, medications or a diagnosed condition as a question to raise with a clinician, never as an instruction to act on. Ask for sources and actually open them. OpenAI's own position, given to Newsweek, is that "ChatGPT is not a doctor and should never be used as a substitute for medical care, diagnosis, or treatment" - this case is a concrete demonstration of why that line is not boilerplate. Reference: DeVerna, M.R., Yan, H.Y., Yang, KC. et al. The invisible risks of AI-generated health information. Nature Health (2026). doi:10.1038/s44360-026-00206-x Source: https://www.newsweek.com/a-man-turned-to-ai-for-diet-advice-his-hospital-stay-exposed-a-blind-spot-12573267
Your answer
Sign in to verify this AI response.
Don't trust us — or the AI. Ask ChatGPT / Ask Claude / Ask Gemini this same question and compare the answers yourself.
More in Nutrition
- AI Eating Disorder Advice High Risk Without Bio — GLM, Gemma and Qwen
- ChatGPT Kidney Diet Micronutrient Numbers Review — ChatGPT (OpenAI)
- Google AI Overview Inspired Taste Fake Banana Bread Mayonnaise — Google AI Overviews and AI Mode
- Chatbot Diet Advice Eating Disorder Ultralow Calorie — AI chatbots
- Four Chatbots Food Safety Advice 40 80 Percent — ChatGPT, Bard, Meta AI, Copilot
- AI Renal Diet Plans Dialysis Energy Potassium Errors — ChatGPT, Copilot, Gemini
- Vegan Vegetarian Diet Chatbots 47 Harmful Answers B12 Iodine — ChatGPT/Gemini/Copilot/DeepSeek/Doubao
- ChatGPT Gemini Perplexity Tpn Plans Critical Safety Errors — ChatGPT, Gemini, Copilot and Perplexity
- Grok Realfood Gov Rectal Food Advice — Grok (xAI)
- AI Chatbots Teen Weight Loss Diet Plans Unbalanced — AI chatbots (multi-model test)
More from this topic
Someone typed a question into an AI assistant that came straight out of a pro-anorexia forum: a weight target, a diet, a training routine stacked on top, and a request for help sticking to it. The assistant answered without a flicker of doubt — guidance the researchers who later graded it filed as high-risk, reinforcing the restriction instead of flagging it. When a team put 100 of those real forum questions to three open-weight models, GLM, Gemma and Qwen, and scored every reply against clinician-written criteria for eating-disorder harm, high-risk answers were roughly twice as common when the model got no user bio to read. Even when the user said outright that they had an eating disorder, safety improved by only 10 to 14 points. So which user is the assistant protecting — me, or the one it assumes I am? Is that safe?
I asked ChatGPT to plan a week of kidney-friendly meals and to list the potassium, phosphorus and sodium in every single day. It answered without hesitation — seven days of meals, each with a precise milligram figure, stated as fact, with no hint that a renal diet is exactly where those numbers have to be right. Would you trust it?
I searched Google for "Inspired Taste banana bread recipe" because I wanted their actual recipe, from them. Google's AI Overview answered at the top of the results with a recipe it presented as Inspired Taste's banana bread, stated as if it were simply their recipe. It wasn't — it differs from the real recipe on their site. Same thing with their homemade mayonnaise: the AI version used less mustard and more lemon juice than their tested recipe, yet it still introduced it as "The highly popular Inspired Taste Homemade Mayonnaise... a fail-proof two-minute recipe." And for their pho, the AI described Inspired Taste's recipe as an "accessible, 'shortcut' version that infuses store-bought beef broth" — while the real recipe starts the broth from scratch and advises against store-bought. I asked again another day and got a different ingredient list, delivered with the same full confidence. So is that recipe actually theirs, or not?