A useful meal plan ai prompt does not produce recipes. Recipes are the easy half and any assistant will give you thirty of them for chicken. The half that takes real planning is deciding what gets cooked on a Wednesday when one person works late, the fridge holds leftovers, and half a jar of dal is already open.
So the prompt below is built the other way round. You supply the constraints first, the plan is generated around them, and the shopping list is generated from the plan rather than written separately. The second and third prompts handle the two common leftovers: an overfull plan and a fridge that is half used.
Plan around what already exists first
List what is in the kitchen before you prompt anything. Half a pack of lentils, a bunch of coriander that will not last, two eggs, and rice in the cupboard constrain the week far more than any preference list. A plan generated before that inventory tends to produce seven new meals and a shopping list that ignores both the food you own and the food you will waste.
Write your constraints in one block and paste them every time. The order matters, because a prompt given preferences without constraints produces a wish list, and a prompt given constraints produces something you will actually cook.
The copy-ready seven-day plan prompt
Build a seven-day meal plan for [number] people using [state your household and any regular guests]. My constraints: [weeknight dinners must be on the table by [time]], [which nights nobody cooks], [maximum total cooking time on weeknights], [equipment I own: pressure cooker, oven, air fryer, blender], [cooking skill level], [what the household will not eat], and [how many leftovers are acceptable, in number of portions]. Already in the kitchen and worth using first: [inventory, including open and half-used items]. Preferred rotation: [meals the household actually eats, at least three]. Foods to avoid entirely: [list]. Produce a table with days Monday to Sunday as rows and columns for dinner, cooking time, what it uses from my existing stock, and what must be bought. Then produce one shopping list grouped by supermarket section, marking each item USES EXISTING STOCK where applicable, and give a total number of items rather than a cost. Then note which two dinners can be swapped without breaking the shopping list. Do not invent nutritional claims or health benefits. Do not count calories or macros unless I give you the figures. Do not repeat a base ingredient as a separate purchase if I already listed it as stock. Do not schedule a meal longer than my stated cooking limit on a weeknight. If a day cannot be filled without breaking a constraint, print WHY NOT and suggest what I would have to change rather than quietly breaking it.
Fill in these details
- Household size and guests: Children, partners, and guests each change portion logic and sometimes the meal itself.
- Time on the table: The real deadline, which is usually earlier than the cooking time suggests because of school runs and bath time.
- Cooking limit: The maximum minutes of active work, not total elapsed time. A pressure cooker is elapsed time; chopping is active time.
- Equipment: Naming what you own is the difference between a usable plan and a plan that assumes an oven.
- Existing stock: Include open packets. Open items are the ones that get wasted.
- Leftover tolerance: Be honest. A plan with no leftovers means cooking every night.
Adapt by deleting sections you genuinely do not need, but keep the refusal lines. The rule about not inventing nutritional claims is the one most often needed, because a meal plan is exactly the kind of request that invites confident health advice the model cannot support.
Turn the plan into one shopping list
Ask for the list in a separate step rather than in the same response. It is shorter, faster to redo, and lets you check the plan against the pantry before committing money to it.
From this meal plan, produce a shopping list: [paste the plan]. Group items by section: produce, protein, dairy, grains and pulses, spices and sauces, frozen, household. Mark items I can substitute, and for each substitution name one alternative of similar cost and similar cooking time. List the items that must be bought fresh this week separately from the items that last longer than a week, because the second group is what gets forgotten. Give the number of distinct items. Do not invent prices, do not suggest shops, and do not add items the plan does not use, including snacks I did not ask for.
The half-used fridge prompt
Run this whenever the plan has slipped or you are a day into the week. It is the most useful of the three for reducing waste.
Here is what I have: [list every open and partial item]. Here is what needs using first: [the two or three items closest to going off]. I can cook for [minutes], I have [equipment], and I am feeding [number]. Give me three options, in order of least effort. For each option: the dish, approximately how long it takes, which of my items it uses up completely, what it adds that I do not have, and whether it keeps for the next day. Put the items closest to going off first in the first option. If an item cannot be used safely in the time available, say so plainly rather than inventing a way to use it, and tell me how to store it to extend its life. Do not assume I have pantry staples I have not listed, and ask me once at the end what staples I actually keep.
Worked example: two cooks, one hour, six constraints
The stated constraints for a household are: two adults and one child, dinner on the table by 19:00 Monday to Thursday, no cooking on Friday, thirty-five minutes of active cooking on weeknights, a pressure cooker and a hob but no oven, one person will not eat mushrooms, and two portions of leftovers are fine.
The generated plan looks like this shape.
| Day | Dinner | Active time | Uses existing stock | Buy |
|---|---|---|---|---|
| Monday | Dal, rice, cucumber salad | 20 minutes | Half jar dal, rice | Cucumber, tomatoes |
| Tuesday | Chana curry, roti | 35 minutes | Coriander, onion | Chana, atta |
| Wednesday | Leftover dal with fried rice | 15 minutes | Monday dal | Spring onion |
| Thursday | Vegetable stew, bread | 30 minutes | Potatoes, onion | Mixed vegetables |
| Friday | No cooking, ordered or eaten out | 0 | ||
| Saturday | Longer cook, project | 90 minutes | Meat or paneer | |
| Sunday | Batch cook for the week ahead | 60 minutes | Grains |
The Wednesday line is the one most generated plans miss. It uses Monday's leftovers, removing a purchase and a cooking session at once, and it is only possible because leftovers were allowed in the constraints. Under the shopping list, the coriander, onion, and potatoes appear once, marked USES EXISTING STOCK, rather than as three separate purchases.
Check the plan before you shop
1. Confirm every meal fits your actual worst night
Look at the two hardest nights, not the average. If Tuesday needs thirty-five minutes and you have thirty on a bad day, swap it with Monday now rather than at 18:30. The prompt tells you which two dinners are swappable, and that column is the one to use.
2. Check the shopping list against your own pantry
Do it by hand, aisle by aisle, before paying. Generated lists are excellent at structure and unreliable about what you already own, especially oil, spices, and staples. The USES EXISTING STOCK markers help, but only if you verified them.
3. Check the perishable load against your real cooking days
If the plan buys fresh produce for four nights but you only cook three, the fourth batch becomes waste by design. Move the perishable meal to a night you will actually cook.
Common meal plan prompt mistakes
- Planning before listing the kitchen. The plan ignores what you own and your shopping list becomes a full restock.
- Ascribing health benefits. Models will confidently attach health claims to dishes. If you have specific nutritional needs, take that to a qualified professional rather than to a prompt.
- Ignoring the worst night. A plan designed for an average evening fails on the evening you planned for.
- Counting leftovers as waste. Planned leftovers that get eaten are the cheapest meal in the week.
Frequently asked questions
How do I stop the shopping list from being too long?
Set an item ceiling in the prompt and require the plan to use stock first. If the list still runs long, the usual cause is seven entirely different dinners, so ask for a rotation of five reusing two base components. Marking long-life items separately also stops you buying weekly what you could buy monthly.
Can these prompts handle allergies and medical diets?
You can state allergies and dietary requirements as hard constraints and ask the plan to respect them, but treat any nutrition or health guidance from a model as unverified. For allergies, cross-check ingredients against the packaging rather than the plan, and take any medical dietary requirement to a qualified professional.
What if my household changes mid-week?
Run the half-used fridge prompt instead of regenerating the plan. Regenerating produces a new ideal week and discards the cooking you have already done, while the fridge prompt keeps what exists and plans forward from it.
Generate the plan, then audit the list by hand
Paste your real constraints and your real inventory, generate the week, and check the shopping list aisle by aisle before you pay. The structure is genuinely useful; the pantry knowledge is yours to supply. For more cooking prompts, browse the cooking prompt collection.