How to read AI-generated health advice critically

General principles such as sleeping enough, eating a variety of foods, and increasing training gradually are well supported. Specific claims such as exact doses, precise percentages, or named studies are where errors and invented details most often appear. If a specific figure matters to your health, check it against a reliable source or a professional.

Run through these quickly before changing what you eat, how you train, or what you take.

• Does it cite a source you can actually open and read?
• Does it match what a qualified professional has told you?
• Does it involve medication, supplements, fasting, or a large change in food or exercise? If so, check with a professional first.
• Does it claim to diagnose or treat a condition? Be wary.
• Is it comfortable saying what it does not know?

Language models generate the most plausible-sounding continuation of a text. They do not look facts up unless a tool is built to make them, and they can blend together studies, invent citations, or state an outdated guideline as current. The output reads as fluent either way, which is what makes errors hard to spot.

Health guidance also changes. Recommendations get revised, and a model trained on older text may repeat a position that has since moved on. When the date of the guidance matters, look for the publication date at a reputable health body or professional society.

Used as a starting point, a chatbot can help you prepare. It can explain unfamiliar terms, suggest questions to ask at an appointment, or summarize a document you paste in. Used as a final authority, it is a poor substitute for a professional who knows your history.

• Ask it to explain, not to decide.
• Ask what would change the answer, and what it does not know.
• Verify anything specific with a reputable source.
• Bring what you learn to a qualified professional if it affects your treatment.

Advice that ignores your history, medications, injuries, allergies, or cycle irregularity may not suit you. A well-designed tool asks about these and lets your answers constrain the output. If it never asks, it is guessing.

For everyday training and meal ideas, AI suggestions are usually low-risk if you apply common sense: build up gradually, stop if something hurts, and keep to foods that suit you. The risk rises when advice involves extreme calorie restriction, very rapid changes, or rules that ignore an injury, pregnancy, an eating disorder history, or a medical condition.

If a plan makes you anxious about food, pushes you to train through pain, or leaves you exhausted, that is a signal to change course and talk to someone.

Health and cycle data is sensitive. Check what a tool collects, whether it shares data with advertisers or third parties, and whether you can export or delete your data. Avoid pasting information into general-purpose chatbots that you would not want stored.

Severe pain, fainting, unusually heavy bleeding, chest pain, disordered eating, or symptoms that are getting worse are not questions for a chatbot. Contact a healthcare professional or emergency services as appropriate.