Correcting a meal scan

No AI reads every plate perfectly on the first try. When a scan doesn't match what's in front of you, NutriShot AI gives you two quick fixes: one for the amount, one for the identification. Both take seconds.

Fix the amount: portion confirmation

If the food is right but the amount looks off, use the portion-size confirmation. After every photo, the app asks whether the portion looks right. Adjust it there and the calories, macros, and micronutrients all recalculate. This step alone takes average calorie accuracy from within 11% of a kitchen scale to within 10%.

Fix the identification: add context

If the AI identified the wrong dish or missed an ingredient, add one or two words of context. A few examples of the kind of hint that works:

• The preparation: "grilled", "fried", "steamed"
• A swapped ingredient: "oat milk", "tofu", "whole wheat"
• The cuisine or dish name when it's ambiguous

The AI takes the hint and corrects its read of the meal. That's by design: a couple of words from you beats a long editing session.

If it keeps missing

For meals that consistently read oddly, try a clearer photo from above with the whole plate in frame, or log by text description instead. If something seems broken, email us. Real examples help us improve the scanner.

• Email support