Nutrition sources & methodology
The data behind Yoked's planned local food catalog.
Updated October 11, 2026
Planned for a Yoked 1.4 update
We are developing a bundled catalog of generic ingredients, foods, and common prepared dishes for Yoked 1.4. This catalog and the local meal workflow described here are planned features. For how currently released versions handle food lookups and Apple Intelligence, see our privacy policy.
USDA FoodData Central
The catalog is derived from USDA FoodData Central, maintained by the U.S. Department of Agriculture's Agricultural Research Service, Beltsville Human Nutrition Research Center. We use these public releases:
- Foundation Foods — April 2026 (JSON archive). Composition data for commodity and minimally processed foods.
- SR Legacy — April 2018 (JSON archive). Historical Standard Reference food composition data, including different preparation states.
- FNDDS 2021–2023 — October 2024 release (JSON archive). The Food and Nutrient Database for Dietary Studies provides nutrients and portion weights for foods and beverages reported in U.S. dietary surveys, including common prepared dishes.
See USDA's dataset downloads and data documentation for release details and descriptions of each data type. Sources accessed October 11, 2026.
License and attribution
USDA identifies FoodData Central data as public domain and publishes them under CC0 1.0 Universal. FoodData Central is the source of the nutrition data; Yoked's catalog adapts the downloaded records for local use and retains their USDA food identifiers and release information.
How the planned local workflow works
On-device AI will interpret a meal description and any supplied amounts to estimate ingredients, preparation state, and portion weight. Yoked will match those foods to the bundled catalog and calculate nutrition from the source values per 100 grams. When available, USDA household portions help convert measures such as cups or pieces into grams. Nutrition values will come from the catalog rather than being invented by the model.
This local meal workflow is designed to calculate nutrition without Private Cloud Compute or a nutrition-server lookup. Other AI features and the handling of meals you choose to save are described in our privacy policy.
The Watch app will also support offline catalog search, portion entry, and local meal logging without the iPhone. Meals saved on Watch will sync to the iPhone when it becomes available. AI meal interpretation remains a separate feature on supported iPhone and Mac devices.
On supported iPhones, meal photos will be interpreted on device and matched to the local catalog. Nutrition-label photos will instead use the readable printed nutrition and serving information. Favorites, Recents, manual entry, and reviewed meals will save locally before any optional device sync or Health export.
Barcode scans and barcode photos will be decoded on the iPhone. A barcode identifies a product; it does not contain its nutrition facts. Previously retrieved product information will be available offline. When saving a reviewed nutrition label, you will also be able to link its per-serving values to a barcode. That association will be remembered only on this iPhone for future offline scans, separately from Open Food Facts data. Scanning or analyzing a label will not save the association until you explicitly save the reviewed food. For an unknown barcode, you can photograph a readable nutrition label, enter nutrition manually, or explicitly choose an online lookup in the crowd-sourced Open Food Facts database. This optional cache is separate from the USDA catalog.
Scope and review
The catalog focuses on generic foods and common dishes. Branded products, restaurant menus, and barcode data are outside its scope; it is not an exhaustive list of every food. Raw and cooked foods can have different nutrition per gram, and recipes vary with ingredients, cooking methods, and water loss. A generic dish represents a reference recipe, not an exact measurement of your meal.
Portions and food matches are estimates. You will be able to review and edit them before saving a meal. A nutrient that is missing from a source record remains unknown rather than being treated as a measured zero.