Useful nutrition estimates start with clear information.

Photos add context. Descriptions fill in gaps.

AI meal analysis uses your typed or spoken description and any meal photo you choose to provide. Describe foods and preparation, and include known oils or sauces. The app can infer foods and portions, but it cannot reliably see every hidden ingredient or determine an exact weight from appearance.

A confidence label or estimate range is a cue to review the result. It is not a measured guarantee of accuracy for your meal. We do not claim a universal percentage accuracy for photo analysis.

Database records and labels provide a starting point.

Simple-food nutrition may be checked against USDA FoodData Central. Barcode lookups use USDA FoodData Central and Open Food Facts. These records can describe different products, serving bases or preparations, so the selected food matters as much as the number.

Nutrition-panel photos produce editable label drafts. A draft must be reviewed against the package before it is used. Label values describe the stated serving, not automatically your whole meal.

Reference: USDA FoodData Central: Data documentation

The amount and preparation determine the calculation.

Matching label values can be scaled to a compatible entered amount. Cooked and dry weights are kept separate: a measured batch yield can provide a conversion, or you can choose a cooked-food estimate.

If you did not weigh the food, say so. An approximate portion from a description or photo is still an estimate even when the package facts are known. Recipe totals also depend on ingredient amounts and how much of the cooking batch you eat.

Unknown does not mean zero.

Some food records and labels do not provide every nutrient. Healthy Beetle preserves unknown values instead of filling them with zero. A known subtotal may therefore be incomplete.

Check the coverage note alongside a total. Enter zero only when the source actually reports zero; leave a missing value blank. A precise-looking decimal does not remove uncertainty in the food match or portion.

You control the review and the save.

Review detected foods, portions and package matches before saving a food entry. Updating the description or label details requires a fresh analysis; changing supported portion amounts can recalculate the displayed nutrition.

AI features ask for consent and send the selected content needed for the feature to OpenAI. Manual food entry remains available. The privacy policy explains what is sent and retained. Nutrition output is general information, not a diagnosis or treatment.