Nutrition Score Optimizer

How to use it

1. 1Copy the prompt into ChatGPT, Claude or Google Gemini.
2. 2Fill the bracketed inputs (product, nutrition panel, scoring system, target).
3. 3Answer its clarifying questions about non-negotiables and constraints.
4. 4Iterate: ask for a version that prioritizes zero taste risk.

See it in action

Score diagnosis for Morningful: sugar content is the anchor holding the score at C; fiber is the under-used favorable lever. Verify against the current Nutri-Score algorithm with your regulatory team before acting.

Levers ranked by score impact:
1. Sugar reduction — the dominant lever; stepwise cuts matter most near threshold points.
2. Fiber increase — raising whole-oat share and adding oat bran lifts the favorable side.
3. Sodium trim — small contribution, low effort.
4. Saturated fat — already low; little to gain.

Taste-risk ranking:
- Low risk: oat bran addition (slight color deepening), sodium trim.
- Medium risk: a staged sugar cut in two moves, holding honey aroma constant.
- High risk: sweetener substitution — excluded by brand policy anyway.

Unlocks at B: stronger signaling at retailers that badge Nutri-Score, eligibility for better-for-you range reviews, and school and workplace channel conversations.

Stepwise path: 1) fiber lift plus sodium trim, re-score; 2) sugar cut step one, bench panel; 3) sugar cut step two, re-score and lock the B before touching anything else.

Want me to draft the bench panel plan for the sugar steps?

The prompt

Here's the start of the prompt. Download the free bundle for the full, ready-to-paste version — plus the installable Claude Skill and Custom-GPT instructions.

# Role
You are a senior nutrition and reformulation strategist for food and beverage companies who knows how threshold-based front-of-pack systems behave — Nutri-Score, Health Star Rating, UK traffic lights. You focus effort on changes that cross a scoring line, and you never trade away the taste that drives repeat purchase.

# Context I'll provide
- Product and category: [PRODUCT]
- Scoring system and market: [SCORING SYSTEM e.g. Nutri-Score in France, HSR in Australia]
- Current nutrition panel: [NUTRITION PANEL — per 100g or 100ml values you have]
- Current and target score: [CURRENT SCORE] to [TARGET SCORE]
- Taste and brand non-negotiables: [NON-NEGOTIABLES e.g. the honey taste, no artificial sweeteners]
- Constraints (optional): [CONSTRAINTS e.g. cost ceiling, ingredient policy, timeline]

# Your task

Frequently asked questions

What is a front-of-pack nutrition score?

Front-of-pack nutrition scores — Nutri-Score in much of Europe, Health Star Rating in Australia and New Zealand, traffic-light labels in the UK — grade a product's overall nutrition on the pack face using an algorithm over nutrients like sugar, sodium, saturated fat, fiber, and protein. Because they are threshold-based, small formulation changes near a cutoff can move the visible grade.

Does it calculate my exact Nutri-Score?

No, and treat any tool that promises to as a starting point only. The skill identifies which levers plausibly move your score and sequences them — but official algorithms are updated over time, so every projected change must be verified against the current published algorithm by your regulatory or nutrition team before you commit to reformulation or on-pack changes.

Which scoring systems does it cover?

Nutri-Score, Health Star Rating, and UK-style traffic lights are handled natively, and the approach adapts to other threshold-based systems if you describe the scheme and market in the prompt. Name the market explicitly — the same recipe can score differently across systems, and the claim rules attached to each score differ by jurisdiction.

Which AI chat tools support this skill?

All the major ones — the prompt runs identically in ChatGPT, Claude, and Google Gemini, and it is model-agnostic by design. Teams working across a portfolio often save it as a Custom GPT or a Claude Skill and run each SKU through it, building a consistent renovation queue ranked by score impact and taste risk.