# What sits behind every label.

FDA rules quoted word for word, USDA data, and AI that only suggests.

[Start free](https://studio.bettermenu.live/signup) [See it in ChatGPT and Claude](https://bettermenu.live/platform/ai-and-agents/)

We send recipe lines only to AI providers that do not train on them and keep none of them after answering.

Figure: A Nutrition Facts panel lifted off the three layers under it: FDA rules, ingredient data and your recipe Three stacked layers in isometric, and a Nutrition Facts panel floating above them. From the top: the FDA rules (21 CFR 101.9), the ingredient data (USDA and your own), and your recipe, in grams. The panel shows a sample 30 g serving: Calories 120, Total Fat 5g, Sodium 35mg, Total Carbohydrate 17g, Protein 3g.

A Nutrition Facts panel lifted off the three layers under it: FDA rules, ingredient data and your recipe

## Five foundations under every label.

- [Compliance engine FDA rules do the math. No AI.](https://bettermenu.live/platform/regulation-engine/)
- [Ingredient data Four USDA sets, plus your specs.](https://bettermenu.live/platform/ingredient-data/)
- [AI and agents Matches, weights, added sugars.](https://bettermenu.live/platform/ai-and-agents/)
- [Security What we store, who sees it.](https://bettermenu.live/platform/security/)
- [How we calculate Six steps, grams to label.](https://bettermenu.live/platform/how-we-calculate/)

Free to start. No card needed.

[Start free](https://studio.bettermenu.live/signup)

[See pricing](https://bettermenu.live/pricing/)

## Where does the AI stop and the compliance engine start?

After your review. AI suggests the USDA matches, any missing weight and, in Studio, the added sugars. None of it reaches the label until you have reviewed it, changed what you need and confirmed it.

Figure: AI suggests, you confirm, the compliance engine calculates: the boundary between the AI and the fixed rules You work in Studio, ChatGPT, Claude. In the AI zone, a guarded box marked "BetterMenu’s AI: no training" and "Zero data retention", the AI reads a recipe line (1 cup pecans) and suggests a USDA match (Pecans, raw). The USDA portion gives its weight (99 g). Grams pass through, ounces and pounds convert by fixed tables, cups and spoons use USDA portions, and AI estimates a weight only when no USDA portion fits. At the boundary you confirm each suggested match before anything crosses. Under it, the compliance engine: fixed rules from 21 CFR 101.9 do every calculation, the rounding and the percent Daily Values, with no AI. The output is the Nutrition Facts label and a saved version with a proof footer. AI never calculates a label value. In ChatGPT or Claude, the chat itself follows that assistant’s terms.

AI suggests, you review and confirm, the compliance engine calculates

## Read next

- [Build Paste a recipe, match USDA data, set servings.](https://bettermenu.live/formulate/build/)
- [Label An FDA-format panel, exported as PDF, PNG or JPEG.](https://bettermenu.live/formulate/label/)
- [Check FDA rounding applied. In Studio, added sugars suggested for you to review and confirm.](https://bettermenu.live/formulate/check/)

## Make your first label today.

No card needed. [See pricing](https://bettermenu.live/pricing/).

[Start free](https://studio.bettermenu.live/signup) [See it in ChatGPT and Claude](https://bettermenu.live/platform/ai-and-agents/)

Source: https://bettermenu.live/platform/