How FoodCheck AI Verdicts Work — and Their Limits

Last updated: 2026-09-24

FoodCheck AI is an AI dietary reference tool that tells you whether a food is safe, caution, or danger for people with a specific health condition. This page explains, plainly, how those verdicts are produced and where they fall short.

How verdicts are produced

The verdicts and summaries on our food guide pages (for example, food × condition pages) come from a food database generated with Google Gemini AI. The generation script asks the gemini-2.5-flash model, prompted to act as a clinical nutritionist, to produce for each food and condition a safe / caution / danger verdict, a 1–2 sentence summary in Korean and English, and the relevant nutrient factors (sodium, purine, potassium, etc.).

When you upload a photo or type a food name, Google Gemini AI generates the verdict at the time of your request. Results for the same food can differ depending on the photo, the preparation, and what you enter.

During generation, the AI was instructed to include only conditions where the food has a meaningful interaction and to skip conditions where it is neutral.

Condition rules used during generation

The rules below are taken directly from the prompt in our database generation script. They are instructions given to the AI — not values measured or calculated for each food.

Diabetes: low GI (<55) = safe; high sugar / refined carbs = danger
Hypertension: >600 mg sodium per serving = danger; >400 mg = caution
Kidney disease: high potassium / phosphorus / protein = danger for CKD stage 3+
Gout: high purine (>150 mg/100 g) = danger; moderate (50–150 mg) = caution
Hyperlipidemia: high saturated fat / cholesterol = danger
Liver disease: high fat / alcohol = danger; moderate = caution
IBS: FODMAP foods = caution or danger; trigger foods = caution
Obesity: high calorie density = caution; >500 kcal per serving = danger
Thyroid: goitrogenic foods (raw cruciferous vegetables) = caution; iodine-rich foods = check
Heart disease: high saturated fat, sodium, trans fat = danger
GERD: acidic, fatty, spicy foods are triggers = caution or danger
Gallstones: high-fat meals are triggers = caution

The generation script contains no numeric rules for gastric cancer, colorectal cancer, breast cancer, pancreatitis, or inflammatory bowel disease.

Limitations

FoodCheck AI verdicts are AI-generated reference information and do not replace medical diagnosis, treatment, or prescriptions.
Dietary guidance differs greatly by type and stage of a condition. For example, chronic kidney disease depends on the stage and whether you are on dialysis, thyroid disease depends on whether it is hyperthyroidism or hypothyroidism, and diabetes depends on your medication and glucose control.
Verdicts assume a typical serving and a common preparation. The actual amount you eat, how it is cooked, and what you eat with it can change the impact.
AI can be wrong. Nutrient details and verdicts may differ from reality, so confirm important decisions with a professional.
The verdicts and explanations on this service have not yet been reviewed by clinicians such as doctors or dietitians.
Always talk to your doctor or dietitian before changing your diet.

Official resources you can check

You can check condition-specific dietary guidance directly with the official organizations below. FoodCheck AI verdicts do not directly cite these resources and have not been reviewed by these organizations.