Before you start: write down the inputs AI cannot safely guess

You need access to a conversational AI tool, a quantity-based food inventory, the number of people eating, the additional grocery budget, cooking-time limits, available equipment, meals eaten away from home, and every allergy, medical diet, religious restriction, or firm dislike. Look in the refrigerator, freezer, and cabinets before you begin. USDA MyPlate recommends the same inventory-first sequence, along with planning for your schedule, leftovers, and a store-section grocery list.

The retained screenshots use this synthetic scenario:

Planning inputValue used in the test
HouseholdTwo adults; breakfast, lunch, and dinner from Monday through Sunday
Additional grocery budgetCNY 260 for meals only
Weekday timeBreakfast 10 minutes; lunch 15 minutes or leftovers; dinner 30 minutes
Weekend timeUp to 45 minutes
PreferencesMild food; no known food allergies in the fixture
EquipmentOne rice cooker and one wok
Existing food1 kg rice, 500 g oats, 8 eggs, 600 g refrigerated chicken breast, 5 tomatoes, 3 potatoes, 2 carrots, 2 onions, 1 head broccoli, 20 frozen dumplings, oil, salt, and soy sauce

Step 1: turn your existing food into structured inventory

What the user does

Check the refrigerator, freezer, and pantry. Record quantities whenever possible. “Chicken breast, 600 g” is useful; “some chicken” forces the model to guess. Ask for organization only so a hidden assumption cannot propagate into an entire menu.

User prompt
I want to plan next week's meals using food I already have.

Please organize the inventory below into a compact table with these
columns: Category, Item, Quantity, and Priority. Use the categories
Staples, Protein, Vegetables, Convenience Food, and Pantry Basics.
Mark perishable foods that should be used first. Do not create a meal
plan yet.

Inventory:
- Rice: 1 kg
- Rolled oats: 500 g
- Eggs: 8
- Chicken breast: 600 g
- Tomatoes: 5
- Potatoes: 3
- Carrots: 2
- Onions: 2
- Broccoli: 1 head
- Frozen dumplings: 20
- Cooking oil, salt, and soy sauce

What the AI returns

ChatGPT converts the raw list into categories and marks the chicken, eggs, tomatoes, carrots, and broccoli as use-first foods. It also exposes one assumption: the chicken is treated as refrigerated rather than frozen.

ChatGPT result organizing rice, oats, eggs, chicken, vegetables, dumplings, and pantry basics into an inventory table with use-first priorities
Step 1 result — the raw pantry list becomes structured inventory, and perishable ingredients receive a visible use-first priority. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.

Result to approve: every food appears once, the quantities match your notes, and the storage assumption is correct. Fix the inventory before continuing if any line is missing.

Step 2: add the constraints that make the menu realistic

What the user does

Supply household size, budget, time, preferences, and equipment. Ask the AI to repeat the brief and identify only missing facts that could change the plan. This creates a visible checkpoint before any meals are invented.

User prompt
Here are my planning constraints:

- Two adults
- Seven days of breakfast, lunch, and dinner
- Additional grocery budget: CNY 260
- Weekday breakfast: 10 minutes maximum
- Weekday lunch: 15 minutes maximum, or leftovers
- Weekday dinner: 30 minutes maximum
- Weekend cooking: up to 45 minutes
- No food allergies; we prefer mild food
- Equipment: one rice cooker and one wok

Please summarize the constraints in a compact checklist, then list
only the missing details that could materially change the plan. Do not
create the menu yet. Keep the response under 180 words.

What the AI returns

The result confirms the planning contract and asks which days are the weekend, whether the chicken is refrigerated or frozen, whether leftover storage is available, whether the budget includes snacks, and whether any meals will be eaten away from home.

ChatGPT checklist confirming household size, seven-day scope, CNY 260 budget, cooking-time limits, food preferences, and kitchen equipment
Step 2 result — ChatGPT repeats the constraints and asks only for details that could materially change the plan. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.

Result to approve: the checklist says what you meant, and every unanswered question receives a real answer. Do not let “no allergies” in this example replace the needs of your own household.

Step 3: generate a first draft—without a shopping list

What the user does

Answer the missing questions, then request one seven-day table. Keep shopping separate: if the menu is weak, an early grocery list only makes the weak assumptions harder to see.

User prompt
Use Monday through Sunday; Saturday and Sunday are the weekend. The
chicken is refrigerated, the dumplings are frozen, and I have
refrigerator and freezer space. We have no additional dislikes. The
CNY 260 budget covers meals only, and no meals will be eaten away from
home.

Create the first draft of the 7-day meal plan. Use a compact table with
these columns: Day, Breakfast, Lunch, Dinner, Active Time, and Existing
Ingredients Used.

Requirements:
- Prioritize the existing ingredients.
- Use leftovers for at least three weekday lunches.
- Keep meal names short.
- Respect all time limits.
- Use Saturday or Sunday for any longer prep.
- Do not create a shopping list yet.

What the AI returns

The draft covers all seven days, includes four leftover lunches, shows active time, and names inventory items. The first and later rows look coherent when read separately.

First rows of a ChatGPT seven-day meal-plan draft with breakfast, lunch, dinner, active time, and existing ingredients used
Step 3 result, part 1 — the first draft appears complete and includes active time plus existing ingredients. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.
Later rows and notes from the first ChatGPT meal-plan draft showing weekend meals and intended leftovers
Step 3 result, part 2 — the draft covers the week, but visible completeness does not prove the quantities work. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.

Result to hold, not approve: the table is visually complete, but it spreads 600 g of chicken across too many meals, expects too much from eight eggs and 20 dumplings, and relies heavily on rice. A good-looking draft is not yet an executable plan.

Step 4: ask the AI to audit its own draft

What the user does

Give ChatGPT an explicit reviewer role and concrete failure modes. Tell it not to revise anything yet, so the original defects remain visible and the repair can be traced to the audit.

User prompt
Do not revise the menu yet. Audit the draft against the original
inventory and constraints.

Check:
- cooking-time limits
- repetition and variety
- whether the listed quantities are sufficient
- ingredients bought for only one meal
- leftover timing
- budget risk
- food-storage or safety issues

Return a compact table with Severity, Issue, Evidence, and Suggested
Fix. Be candid and keep it under 250 words.

What the AI returns

The audit identifies three high-severity problems: the chicken allocation is impossible and leaves refrigerated chicken too late in the week; eight eggs are unlikely to cover the proposed meals for two adults; and 20 dumplings are stretched across too many servings. It also flags rice quantity, repetitive meals, unpriced additions, and undefined leftover portions. The time limits mostly pass.

ChatGPT audit table flagging chicken, egg, dumpling, rice, repetition, budget, and leftover problems in the first menu draft
Step 4 result — the explicit audit catches three high-severity quantity problems and several medium-risk execution gaps. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.

Result to approve: the audit names the issue, evidence, and correction—not merely “add more variety.” You still decide which fix fits your household.

Step 5: revise the menu with explicit corrections

What the user does

Convert the audit into requirements the next version must preserve. This is more reliable than asking for a generic “better” menu.

User prompt
Revise the menu using the audit.

Required corrections:
- Cook all 600 g of chicken on Monday as four servings; eat two at
  dinner and use two for Tuesday lunch.
- Remove chicken from every later meal.
- Use the 20 dumplings in no more than two meals.
- Add low-cost tofu, beans, noodles, bread, milk, and seasonal fruit
  where useful.
- Add eggs to the shopping list later if the eight existing eggs are
  not enough.
- Create at least three safe leftover lunches.
- Make Friday dinner a flexible “use-it-up” meal.
- Keep weekday meals within the original time limits.
- Avoid buying an ingredient for only one dish.

Return the final seven-day menu as a compact table with Day, Breakfast,
Lunch, Dinner, Active Time, and Batch/Leftover Note. Do not generate the
shopping list yet.

What the AI returns

The new menu uses all chicken by Tuesday, limits dumplings to Monday and Saturday, and creates four planned leftover lunches. Tofu, beans, noodles, bread, milk, and fruit support several meals instead of becoming one-use purchases.

Revised ChatGPT seven-day menu with corrected chicken and dumpling allocation plus batch and leftover notes
Step 5 result — the revised menu gives every batch a destination and removes the impossible later-week chicken meals. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.

Result to approve: each four-serving dinner states which two servings become lunch. “Leftovers” without a source meal, quantity, storage plan, and day is still an unresolved instruction.

Step 6: create one consolidated shopping list

What the user does

Generate the list only after the menu is stable. Require the model to deduct the original inventory, merge repeated ingredients, group the result by store section, and flag uncertain pantry items.

User prompt
Based only on the final menu, create a consolidated shopping list for
two adults. Deduct the original inventory first.

Group additions into:
- Produce
- Protein and dairy
- Staples
- Pantry items

For each item, give the suggested quantity, the meals it supports, and
one practical substitute. Do not repeat items by day. Do not estimate
prices yet. Flag anything that might already be covered by the pantry
basics.

What the AI returns

The result explains which meals use each addition and offers practical substitutions, such as lentils for beans or extra rice for noodles. Existing oil, salt, and soy sauce are not purchased again.

ChatGPT consolidated shopping list showing produce and protein additions with quantities, supported meals, and substitutes
Step 6 result, part 1 — produce, protein, and dairy additions are consolidated after the original inventory is deducted. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.
ChatGPT shopping list showing staple and pantry additions, possible pantry-covered items, and practical substitutions
Step 6 result, part 2 — repeated ingredients are combined once, and uncertain pantry items are flagged for a manual check. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.

Result to check manually: confirm household package sizes, how much fruit people will actually eat, and whether “soup seasoning” is necessary. Delete anything already available before pricing the list.

Step 7: test the list against the budget

What the user does

AI does not automatically know the current price at your store. Ask for transparent illustrative assumptions and explicit arithmetic, or replace the estimates with current prices from your own store, app, flyer, or receipt.

User prompt
Add a budget check to the shopping list.

Use clearly labeled illustrative CNY prices, not claims about current
local prices. For each item, show Quantity, Estimated Unit Price, and
Subtotal. Then calculate:
- estimated total
- CNY 260 budget
- remaining amount
- whether a 10% price buffer still fits

If the buffer does not fit, suggest the smallest substitutions first.
Keep the math explicit and round to whole CNY.

What the AI returns

The illustrative list totals CNY 107. Ten percent is CNY 10.70 and is rounded to CNY 11, so the buffered total is CNY 118. Against CNY 260, CNY 142 remains unallocated.

ChatGPT budget table with illustrative CNY quantities, estimated unit prices, and subtotals for the shopping list
Step 7 result, part 1 — each estimate is visible, labeled illustrative, and tied to one shopping-list quantity. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.
ChatGPT budget check showing CNY 107 estimated spend, CNY 11 buffer, CNY 118 buffered total, and CNY 142 remaining
Step 7 result, part 2 — the arithmetic is explicit: CNY 107 estimated, CNY 118 with the rounded buffer, and CNY 142 unallocated. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.

Result to approve only after local verification: every quantity has a price, every subtotal can be recalculated, the total matches the rows, and the buffer still fits after replacing illustrative figures with real ones. A plan that spends the exact limit has no room for package-size or price differences.

Step 8: assemble one final execution pack

What the user does

The correct information is now spread across several messages. Ask the AI to consolidate only the approved versions. Explicitly prohibit a fresh menu or new budget arithmetic.

User prompt
Assemble the confirmed work into a compact, print-ready weekly
execution pack. Do not change the approved menu or the budget math.

Include:
1. A summary showing 2 adults, 7 days, CNY 260 budget, ¥107 estimated
   spend, and ¥118 with buffer.
2. The final menu with Day, Breakfast, Lunch, and Dinner.
3. A categorized shopping checklist with quantities.
4. A short prep schedule.
5. A leftover map for Tuesday through Friday.
6. A note identifying which prices must be checked locally.

Keep it practical, skimmable, and under 750 words.

What the AI returns

The final response is one weekly execution pack. The five frames below document its summary, complete menu, shopping checklist, prep schedule, leftover map, and unresolved local-price check.

Final ChatGPT weekly execution pack summary for two adults, seven days, CNY 260 budget, and CNY 118 buffered estimate
Step 8 result, part 1 — the final pack begins with one household, time, and budget summary. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.
Final categorized ChatGPT shopping checklist with produce, protein and dairy, staples, and pantry items
Step 8 result, part 3 — the final checklist keeps quantities grouped by store section. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.
Final ChatGPT preparation schedule showing Monday chicken batch cooking, midweek prep, and the start of the leftover map
Step 8 result, part 4 — the prep schedule connects batch cooking to the following lunch rather than leaving leftovers vague. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.
Final ChatGPT leftover map for Tuesday through Friday plus a reminder to verify local grocery prices
Step 8 result, part 5 — every weekday leftover names its source meal, and the locally changing prices remain unresolved. Credit: Grow AI Skills editorial ChatGPT test capture. Open full-size image.

Final result: the pack preserves the approved menu and arithmetic rather than silently generating another version. It remains a planning draft until you verify ingredients, allergies, food safety, nutrition needs, and local prices.

The finished seven-day menu

DayBreakfastLunchDinner
MondayMilk oats + fruitDumplings + broccoliChicken-broccoli rice
TuesdayEgg + toast + fruitLeftover chicken riceTofu-tomato noodles
WednesdayMilk oats + fruitLeftover tofu noodlesBean-potato rice
ThursdayEgg + toast + fruitLeftover bean-potato riceTofu-carrot noodles
FridayMilk oats + fruitLeftover tofu noodlesUse-it-up fried rice
SaturdayEgg-tomato toastDumpling vegetable soupBean-tofu rice bowls
SundayFruit oats + milkTofu-vegetable noodlesPotato-bean rice bowls

Shopping checklist after deducting the inventory

SectionAdditions
ProduceAbout 2 kg seasonal fruit; about 750 g mixed seasonal vegetables; 1 bulb garlic
Protein and dairyAbout 800 g firm tofu; 500 g dried beans; 1 L milk; 6 additional eggs
StaplesAbout 500 g noodles; 1 loaf bread, about 400–500 g; 1 kg additional rice
Pantry1 small pack soup seasoning or stock, optional; omit if salt and soy sauce are sufficient

Illustrative budget

ItemQuantityIllustrative priceSubtotal
Seasonal fruit2 kg¥8/kg¥16
Mixed vegetables750 g¥10/kg¥8
Garlic1 bulb¥3/bulb¥3
Firm tofu800 g¥8/500 g¥13
Dried beans500 g¥12/500 g¥12
Milk1 L¥13/L¥13
Eggs6¥1.5/egg¥9
Noodles500 g¥8/500 g¥8
Bread1 loaf¥12/loaf¥12
Rice1 kg¥8/kg¥8
Optional soup seasoning1 pack¥5/pack¥5
Estimated total¥107

These are tutorial estimates. Replace all of them with local figures before relying on the total.

Prep and leftover schedule

  • Monday: Cook all 600 g of chicken as four servings. Eat two, promptly refrigerate two for Tuesday lunch, and cook extra rice.
  • Tuesday: Make four servings of tofu-tomato noodles and reserve two for Wednesday lunch.
  • Wednesday: Make four servings of bean-potato rice and reserve two for Thursday lunch.
  • Thursday: Make extra tofu-carrot noodles and reserve two servings for Friday lunch.
  • Friday: Use remaining rice, vegetables, tofu or beans, and eggs in the flexible dinner.
  • Weekend: Use the longer cooking window for beans, vegetable prep, and Sunday components.

Reuse the workflow with your own ingredients

This single starter prompt tells the AI to stop after structuring your inputs. Continue with the same audit, revision, list, budget, and consolidation prompts from the steps above.

Reusable starter prompt
Help me create a practical seven-day meal plan.

Household:
- People: [number]
- Meals to plan: [breakfast/lunch/dinner]
- Dietary requirements or allergies: [details]
- Foods we dislike: [details]

Existing food and quantities:
[paste inventory]

Constraints:
- Additional grocery budget: [amount and currency]
- Weekday breakfast time: [minutes]
- Weekday lunch time: [minutes or leftovers]
- Weekday dinner time: [minutes]
- Weekend cooking time: [minutes]
- Available equipment: [equipment]
- Meals eaten away from home: [details]
- Leftover storage available: [yes/no]

First, organize the inventory and summarize the constraints. Ask only
for missing information that could materially change the plan. Do not
generate the menu until I answer.

After I confirm the inputs, work in this order:
1. Generate a first-draft seven-day menu.
2. Audit quantities, time, repetition, leftovers, storage, and budget.
3. Revise the menu using the audit.
4. Deduct my inventory and create a consolidated shopping list.
5. Add transparent price estimates and a budget buffer.
6. Assemble the confirmed work into one execution pack.

Label every assumption and do not present estimated local prices as
current facts.

Fix the most common failure modes

ProblemNext instruction to give the AI
The menu is too complicatedLimit weekday dinners to one wok, one pot, or a defined number of active steps. Separate active time from unattended rice-cooker time.
The shopping list is too longIdentify ingredients used in only one meal and replace them with ingredients already used elsewhere.
The menu feels repetitiveChange the meal format rather than every ingredient. The same tofu can support noodles, rice bowls, and fried rice.
The local total exceeds the budgetRank substitutions by savings and disruption, then replace the most expensive nonessential item first.
The week changesKeep one use-it-up meal and move any perishable ingredients from the skipped meal into that slot.
A leftover has no destinationName its source meal, number of servings, storage container, refrigerate-by time, and exact day of use.

The workflow to remember

Inventory → constraints → first draft → audit → revision → shopping list → budget → execution pack.

Each stage produces something you can inspect before the next stage depends on it. That is the difference between accepting a confident weekly table and building a plan whose quantities, leftovers, shopping list, and arithmetic you can actually check.