Prompt
You are helping a registered dietitian draft a meal-pattern scaffold I will fully verify before any client sees it. Build a weekly {{eating_pattern}} framework and adapt it around these constraints: {{restrictions}}.

Structure I want: {{structure}}.

Hard rules:
- Do NOT invent calorie, macronutrient, or micronutrient numbers. Where a target or total would normally go, write [RD TO CALCULATE from food-composition data] instead of guessing.
- Treat every listed restriction or allergy as an absolute exclusion — no exceptions, no "small amounts."
- Do not tailor this to a diagnosis, medication, or lab value. If a choice would depend on one, write [MNT — RD DECISION] and leave it to me.
- Offer variety and realistic swaps, but flag any ingredient that commonly hides an allergen (e.g., soy in sauces, gluten in seasonings).

Output as a day-by-day pattern (meal slots and example foods) plus a short swaps list. No health claims about what the pattern will cure or treat.

Fill in your details and the prompt updates live — then copy.

What you get back (excerpt)

Monday - Breakfast: Greek yogurt, berries, gluten-free oats. Target: [RD TO CALCULATE from food-composition data] - Lunch: chickpea and roasted-veg grain bowl (use certified GF grains) - Snack: apple + peanut butter - Dinner: lentil and tomato stew, side salad Swaps: sub tofu for chickpeas; sub quinoa for rice. Allergen flags: check that vegetable broth and any seasoning blends are certified gluten-free. Weekly totals and any condition-specific targets: [MNT — RD DECISION].

The full workflow

  1. Give the model only general pattern and restriction inputs — no client identity, diagnosis, or labs
  2. Generate the scaffold, then recalculate every nutrient and portion in a validated food-composition tool
  3. Resolve each [RD TO CALCULATE] and [MNT — RD DECISION] flag with your own clinical judgment
  4. Deliver the finished plan through your practice software, not from the chatbot thread

Watch out for

Verify every number: AI mis-estimates nutrients badly — studies found Vitamin D underestimated by 100%, potassium by ~50%, and energy needs set ~695 kcal too low. Following its raw numbers is unsafe, especially for diabetes, CKD, allergies, or any condition where precision matters.

Scope: an individualized plan built around a medical condition is medical nutrition therapy — the domain of a registered dietitian, and in title-protection or licensure states not something an unlicensed 'nutritionist' may provide. Check your state's law.

Privacy: describe the pattern generically. Never paste a client's name, condition, or lab results into a consumer AI tool to 'personalize' a plan.

Where this comes from

Every use case on this site is grounded in real reports from working nutritionists — not invented by us.

More AI use cases for nutritionists

← All 6 use cases: How Nutritionists Use AI