How to use AI for meal planning as a coach. The six-step workflow, six prompts, and the line the model never crosses.
AI is good at one part of meal planning: turning a clear brief into a structured 7-day draft in two minutes. It is bad at everything that makes the plan safe and personal: it guesses portions, it does not know the client, and it will write a medical claim if you let it. This page is the workflow that keeps those two facts apart. Intake data stays in your system, a neutral brief goes to the model, the draft comes back, and you run the swaps, the allergen check and the sign-off before a client sees a line.
By Markus Evers · Published 6 September 2026
the short version
Use AI for meal planning in six steps: keep intake data in your system, write a neutral brief with targets and constraints but no identity, get a 7-day draft, add swap rules, run the allergen and safety check against your own client record, then verify totals and sign off. Below: the six prompts and the safety limits.
What is the workflow for using AI for meal planning as a coach?
The workflow for using AI for meal planning as a coach has six steps: intake data in your own system, a neutral brief, the draft, swap rules, the allergen and safety check, and coach review with delivery. The model touches three of the six. The other three, including the safety check, are yours, and that split is what makes the workflow usable with real clients.
| Step | What you do | What the model does |
|---|---|---|
| 1. Intake data, kept in your system | Targets from your TDEE and macro calculation, food preferences, dislikes, budget, cooking time, equipment, meals per day, and every allergy and medical flag, all stored in the client's record. | Nothing. The model is not part of this step. |
| 2. Neutral brief | Translate the record into a brief with no identity: "1,900 kcal, 150 g protein, 3 meals + 1 snack, 25 minutes cooking max, no dairy, no shellfish, budget-conscious, likes eggs, rice, chicken thighs, lentils". | Receives the brief only. |
| 3. Draft | Prompt 1. Ask for a 7-day table with per-meal macros, a repeat-friendly structure, and a list of every portion the model guessed. | Writes the draft to the targets. |
| 4. Swap rules | Prompt 2. Ask for like-for-like swaps per meal so the client has choice without breaking the totals. | Writes the swap list. |
| 5. Allergen and safety check | Against the real client record, not the brief: every ingredient on the plan and in the swaps checked against allergies, intolerances, medication and medical flags. Anything uncertain is removed. | Nothing. This step is yours. |
| 6. Coach review and delivery | Verify totals in a food database or your meal-planning software, replace anything the client would not cook, write the summary in your voice, and deliver the plan where the client logs food. | Prompt 5 drafts the summary; you rewrite it. |
Step 1 assumes you already have the targets. If not, set them first with how to calculate TDEE and macros for clients; the manual version of steps 2 to 6, without a model, is how to build a client meal plan.
What do you paste into the model, and what stays in your system?
Paste the shape of the brief and keep the person: targets, constraints and food preferences go in, while names, health data, allergies tied to an identity and any verbatim client message stay in your coaching system. Health data is a special category under GDPR, and a public chat window is not a place it belongs. "No dairy, no shellfish" as a menu constraint with no identity attached is a brief; a named client's allergy list is a record.
| Paste in (the brief) | Keep in your system (the person) |
|---|---|
| Calorie and macro targets you calculated | The client's name, email, photos, weight history |
| Meals per day, cooking time, equipment, budget | Medical conditions, medication, diagnoses |
| Liked and disliked foods, as a plain list | Allergies and intolerances tied to a person (describe as menu constraints only: "no dairy") |
| Cuisine and cultural preferences | Check-in messages, intake forms, anything pasted verbatim |
| The structure you want back (table, columns, units) | Anything that would identify the client if the chat leaked |
The wider rules for handling client information, including what your privacy notice should say about AI tools, are in client data privacy for online coaches.
Which six prompts cover the meal-planning workflow?
Six prompts cover the workflow: a 7-day draft from targets, a swap list, a budget version, a batch-cook version, a client-facing summary and a cultural adaptation. Every one asks the model to return a "Portions I estimated" list, because the guessed gram amounts are where AI meal plans go wrong, and every one bans medical claims. Fill the [YOUR ...] fields from the neutral brief, never from the client record.
Draft a 7-day plan from targets
When to use it: Step 3. You have the neutral brief and want the first structured draft.
What to paste in: Targets, meal count, cooking constraints, liked and disliked foods, cuisine, budget. No identity.
You are helping me, a nutrition coach, draft a 7-day meal plan for a healthy adult client. I will check every number and sign off on the plan; your job is a structured first draft, not advice. TARGETS (set by me, do not change them) Calories per day: [e.g. 1,900 kcal] Protein: [e.g. 150 g] Carbohydrate: [e.g. 190 g] Fat: [e.g. 60 g] Meals per day: [e.g. 3 meals + 1 snack] CONSTRAINTS Cooking time per meal: [e.g. 25 minutes max, 45 on Sunday] Equipment: [e.g. hob, oven, no air fryer] Budget: [e.g. budget-conscious, supermarket own brands fine] Foods to build around: [e.g. eggs, chicken thighs, rice, lentils, oats, Greek-style yoghurt alternative] Foods to leave out: [e.g. dairy, shellfish, coriander] Cuisine or style: [e.g. simple Scandinavian home cooking] Repeats: [e.g. breakfast may repeat all week; lunches and dinners at most twice] OUTPUT 1. A table with columns: Day, Meal, Dish, Ingredients with gram amounts, Calories, Protein, Carbs, Fat. 2. A daily totals row for each day, within 5% of the targets. 3. A list titled "Portions I estimated" naming every line where you guessed an amount or used a generic value, so I can verify it. 4. A 10-item shopping list grouped by aisle. RULES No medical claims, no supplement advice, no language about weight loss speed. Plain dish names a client would recognise. Metric units. Do not add motivational text.
Swap list for every meal
When to use it: Step 4. The draft is checked and you want the client to have choice without breaking the totals.
What to paste in: The finished draft table from prompt 1.
Here is a 7-day meal plan draft I have already checked. For every lunch and dinner, give me two like-for-like swaps that keep the meal within 50 kcal and 5 g protein of the original. A swap must use the same cooking time or less and the same constraints (no [e.g. dairy, shellfish]). Present it as a table: Original dish, Swap A with gram amounts and macros, Swap B with gram amounts and macros. Add a "Portions I estimated" list at the end for anything you guessed. Do not change breakfasts or snacks. No commentary. The plan: [PASTE THE TABLE]
Budget version
When to use it: A client's circumstances changed, or the first draft came back too expensive.
What to paste in: The checked draft plus the weekly food budget.
Rewrite this 7-day meal plan for a weekly food budget of about [e.g. EUR 45] for one person, shopping at a standard supermarket in [COUNTRY]. Keep every day within 5% of the original calories and protein. Prefer eggs, tinned fish, frozen vegetables, oats, rice, lentils, chicken thighs and own-brand products. Keep the same constraints (no [e.g. dairy, shellfish]). Output the same table format, a daily totals row per day, a shopping list with an estimated price per line and a total, and a "Portions I estimated" list. Do not use the words "cheap" or "diet" anywhere. The plan: [PASTE THE TABLE]
Batch-cook version
When to use it: A client who will only follow a plan if it takes one cooking session on Sunday and one midweek.
What to paste in: The checked draft plus the two cooking windows and the fridge/freezer situation.
Turn this 7-day meal plan into a batch-cook plan with two cooking sessions: [e.g. Sunday, 90 minutes] and [e.g. Wednesday, 45 minutes]. Keep each day within 5% of the original calories and protein and the same constraints (no [e.g. dairy, shellfish]). Output: 1. A cooking session plan: what is cooked, in what order, with oven and hob running in parallel where sensible. 2. A container list: what goes in each container, portion size in grams, fridge or freezer, and which day it is eaten. 3. The daily eating table with the same columns as the original. 4. A "Portions I estimated" list. Assume [e.g. 6 fridge containers, freezer available]. Do not include meals that go off before they are eaten. The plan: [PASTE THE TABLE]
Client-facing summary
When to use it: Step 6. The plan is checked and you want a short note to send with it that you will rewrite in your own voice.
What to paste in: The checked plan and the two or three things you want the client to notice. No client name.
Write a short note to go with this meal plan, which I will edit before sending. Plain words, second person, 90 to 130 words. Structure: what the week is built around (one sentence), the two things I want them to notice [e.g. breakfast repeats on purpose; every dinner has a swap], what to do if a day goes wrong (one sentence: eat the next meal as planned), and one question to answer at the next check-in. No "journey", no "fuel", no "nourish", no exclamation marks, no emoji, no medical or weight-loss claims. Two versions: A neutral, B warmer. Two messages I wrote to clients before, for tone: [PASTE] The plan and my notes: [PASTE]
Cultural adaptation
When to use it: The draft is nutritionally right but the food is not what the client eats at home.
What to paste in: The checked draft, the cuisine, and a few dishes the client actually cooks.
Adapt this 7-day meal plan to [e.g. Turkish home cooking as made by a working parent on weeknights]. Keep every day within 5% of the original calories and protein and the same constraints (no [e.g. dairy, shellfish]). Use dishes like [e.g. menemen, mercimek corbasi, bulgur pilav, tavuk sote] and standard ingredients from a [COUNTRY] supermarket or a local grocer. Keep the cooking times. Output the same table with gram amounts and per-meal macros, daily totals, a shopping list, and a "Portions I estimated" list that names every dish where you converted a traditional portion into grams, so I can verify it. Do not caricature the cuisine; ordinary weeknight food. The plan: [PASTE THE TABLE]
For the rest of a nutrition coach's week, from check-in replies to grocery lists, the 30 ChatGPT prompts for nutrition coaches are grouped by job and follow the same never-paste-health-data rule. The batch-cook prompt works best when the client already runs a prep routine; how to set one up is in meal prep for clients.
Can an AI diet planner write a client's plan?
It can write the draft; it cannot write the plan. An AI diet planner returns a week of meals that hit the numbers on paper, with portions estimated, allergies unknown and no idea what this client will actually cook. The seven-check review below turns that draft into coaching, and the table shows where each check bites. The consumer side, an AI calorie counter in the client's pocket, is covered in what AI calorie counters get wrong.
| What the planner does | What it misses | What the coach adds | The check |
|---|---|---|---|
| Drafts a week of meals at the targets | Allergies, medication and conditions it was never told | The neutral brief and the medical check against your own record | Allergen and medical check |
| Hits the calorie and macro numbers on paper | Portion estimates that are wrong by 100 to 300 kcal | Verified totals from a database, not the model's guess | Verified totals |
| Suggests recipes and swaps | Whether this client will cook any of it on a Tuesday | Realism for this client's kitchen and week | Every meal realistic |
| Writes confident explanations | Claims it cannot back, and a tone that is not yours | No claims; the summary in your voice | No claims, your voice |
What are the safety limits for AI meal planning?
The safety limits for AI meal planning are four: no medical claims in any plan or note, healthy adults only with every therapeutic case referred to a dietitian or doctor, every number verified by the coach before delivery, and the coach signs off on every plan. A model will breach all four if the prompt lets it, which is why each prompt above carries the limits inside it.
01
No medical claims
Nothing about hormones, conditions, cures or weight-loss speed.
02
Healthy adults only
Diagnosed conditions, eating disorders, pregnancy and therapeutic diets go to a dietitian or doctor.
03
Every number verified
The model guesses portions. Totals are checked in a database or your software.
04
The coach signs off
No plan reaches a client without your review and your name on it.
How do you review an AI meal plan before the client sees it?
Review an AI meal plan with seven checks that take 10 to 15 minutes: verified totals, the allergen and medical check against your own record, referrals made where needed, every estimated portion replaced, every meal realistic for this client's kitchen and week, no claims, and a summary in your own voice. The first draft saved you the afternoon; the review is where the plan becomes coaching. The same caution applies to what the client logs: what AI calorie counters get wrong has the 4-method error table and the three lines to tell a client who uses one.
- 01
Daily totals verified in a food database or meal-planning software, not taken from the draft.
- 02
Every ingredient on the plan and in the swaps checked against the client's allergies, intolerances, medication and medical flags in your own record.
- 03
Any client with a diagnosed condition, an eating disorder history, pregnancy or a therapeutic diet referred to a dietitian or doctor; no AI plan.
- 04
Every "Portions I estimated" line replaced with a weighed or database value.
- 05
Every meal is something this client would actually cook in the time they have.
- 06
No medical, hormonal or weight-loss-speed claim anywhere in the plan or the note.
- 07
The summary is rewritten in your voice, with the client's name and their actual week in it.
Where the checked plan goes. A draft in a chat window is not a plan a client can follow. Coachway builds meals against per-client calorie and macro targets from a library of 1,100+ recipes and 3,900+ ingredients, auto-scales portions to the target and writes the shopping list, in the same place the client logs food and checks in. Use the model for the sketch, your platform for the plan: EUR 69 per month for up to 5 clients plus EUR 9 per additional client, every feature included, 14-day free trial.
Frequently asked questions about using AI for meal planning.
How do I use ChatGPT to make a meal plan for a client?
Give the model a neutral brief, not the client: the calorie and macro targets, the number of meals, the foods the client likes and avoids described without a name, the budget, cooking time and equipment. Ask for a 7-day draft in a fixed table format with per-meal macros. Then run the swap list, do the allergen and safety check against your own client record, verify the totals, and rewrite the client-facing summary in your voice. The six prompts on this page cover each step.
Is it safe to put client information into ChatGPT for meal planning?
Only the neutral shape of the brief. Never paste a client's name, email, weight history, medical conditions, medication, allergies, or messages into a public AI tool; health data is a special category under GDPR. Describe constraints without identity ("no dairy, no shellfish, 1,900 kcal, 150 g protein") and keep the real record in your coaching system. The allergen check is done by you, against that record, after the draft exists.
Can AI calculate calories and macros for a client?
It can estimate, and it is often wrong by 10 to 20 percent on a meal, so treat every number in a draft as a placeholder. Set the targets yourself from a TDEE calculation, ask the model to build to those targets, then verify the totals in a food database or your meal-planning software before the plan reaches the client. The model does not weigh food; it guesses portions.
What should a coach never let AI do in nutrition coaching?
Never let it diagnose, treat or manage a medical condition, prescribe a plan for someone with an eating disorder, a diagnosed gut condition, diabetes, pregnancy or a supplement-medication question, or write a claim like "this plan will fix your hormones". Those clients belong with a registered dietitian or doctor. AI drafts menus for healthy adults; the coach signs off on every plan.
How long does it take to build a meal plan with AI?
About 20 to 30 minutes for a plan that used to take an afternoon, most of it in review: 5 minutes writing the brief, 2 minutes for the draft, 10 to 15 minutes checking macros, swaps and allergens, and 5 minutes rewriting the summary in your own voice. The saving is in the first draft. The review time does not shrink and should not.
Should I send the AI-generated meal plan straight to the client?
No. A draft leaves the model in a generic voice with unverified numbers and no knowledge of the person. Check every total, make sure nothing on the plan conflicts with what you know about the client, replace any meal they would not actually cook, and write the summary yourself. Clients can tell when a plan was not built for them, and that is the moment they stop following it.
How do I make an AI meal plan fit a client's culture or cuisine?
Say so in the brief and give examples: "meals should be Turkish home cooking a working parent makes on weeknights; include menemen, mercimek corbasi, bulgur pilav". Ask the model to keep the targets and swap the food, not the other way round. The cultural adaptation prompt on this page does that, and asks the model to flag any dish where it guessed the portion so you can check it.
What is the best AI tool for meal planning as a coach?
For the draft, any capable model works: ChatGPT, Claude or Gemini produce a similar 7-day table from the same brief. The difference is what happens after the draft. Coaching software with a recipe library, per-client targets and auto-scaling portions turns a checked draft into a plan the client can log against, which a chat window cannot. Use the model for the sketch and your platform for the plan.
Can an AI diet planner write a client's meal plan?
It writes the draft: a week of meals that hits the calorie and macro targets on paper. It does not know the client's allergies, medication or conditions unless you break the data rule to tell it, its portions are estimates, and it has no idea what the client will cook. The seven-check review on this page (verified totals, the allergen and medical check, referrals, portions, realism, no claims, your voice) is what turns the draft into a plan.
The model writes the draft in two minutes. The coach makes it safe, personal and followed. Keep those two jobs separate and AI meal planning is worth the afternoon it saves.
Keep reading
all guidesHow to Create a Meal Plan for a Client (2026 Method)
Building a client meal plan is an ordered method, not a meal-by-meal script: set the calorie and macro frame first, map the client's real food life, choose the structure - fixed, swappable menu or macro targets - then build meals as options with swaps and adjust from check-ins. The methodology behind a plan that gets followed, not a generic PDF. General information, not medical or dietetic advice.
Read the guide30 ChatGPT Prompts for Nutrition Coaches (Copy-Paste 2026)
Thirty copy-paste ChatGPT prompts for nutrition coaches, grouped by job - meal ideas, recipe swaps, grocery lists, client education and check-in replies - each scoped to coaching healthy adults, never a medical condition.
Read the guideHow to calculate TDEE and set macros for coaching clients (2026)
How to calculate TDEE and set macros for coaching clients: the formula, activity multipliers, a worked example, and how to adjust from real check-in feedback. General information, not medical or dietetic advice.
Read the guideFree tools for this: TDEE Calculator · Calorie Surplus Calculator · Calories Per Meal Calculator
Ready-made templates: AI Prompt Pack for Online Coaches · Meal Plan Template
Where the checked plan becomes the client's plan
Per-client targets, a recipe library that auto-scales to them, food logging and weekly check-ins on one screen, with payments through your own Stripe. Start the 14-day free trial and move one client over: cancel any time under Billing.