- Why I even bothered with this
- The prompts I use, in order
- A real week, with real numbers
- The uncomfortable bit nobody selling this trend wants to say
- Adjusting the prompts for different goals
- Family version of the prompt
- Mistakes I made so you don't have to
- A shorter version if you're in a rush
- Frequently asked questions
- Sources worth reading
The short version: ChatGPT can build you a five-day meal plan and a shopping list sorted by supermarket aisle in about ninety seconds, but only if you give it your actual constraints (budget, fridge contents, how much time you have) rather than a vague "help me eat healthier." I'll give you the exact prompts I've used every Sunday since January, plus what goes wrong when people skip the boring detail work.
Why I even bothered with this
I used to spend Sunday mornings scrolling recipe sites with three tabs open, half-writing a shopping list on the back of an envelope, and still ending up at Tesco on a Wednesday buying the same chicken thighs I already had in the freezer. It wasn't a big problem. It was just a slow, annoying one, repeated fifty-two times a year. If you cook, cater or run a food business and have a lesson worth sharing, you can write for us about food.
In January I started running the whole thing through ChatGPT instead. Not because I thought it would change my life, but because I was curious whether it could save the twenty minutes of dithering that came before the twenty minutes of writing a list. It did, most weeks, once I stopped giving it lazy prompts and started giving it real information.
The prompts I use, in order
These aren't theoretical. This is the sequence, copied more or less word for word from my own chat history.
1. The meal plan prompt
"I need five dinners for this week for two adults. Budget is around 45 pounds total for the ingredients. We have chicken thighs, a bag of frozen peas, rice, and half a jar of pesto already in the house, so try to use those. No shellfish, keep two meals vegetarian, and nothing that takes longer than 35 minutes to cook on a weeknight. List each meal with a one-line description, not a full recipe yet."
The specificity matters more than the wording. Budget, what's already in the fridge, dietary limits, and a time cap. Leave any one of those out and you get generic suggestions that ignore your actual kitchen.
2. The full recipe prompt
"Now give me the full recipe for meal 3, with exact quantities for two people, and tell me which bits can be prepped on Sunday and stored so the weeknight cook time is under 15 minutes."
I only ask for full recipes on the meals I don't already know how to make. If it's spaghetti bolognese, I don't need ChatGPT to tell me how to brown mince.
3. The shopping list prompt
"Turn these five meals into one combined shopping list. Group it by supermarket section: fresh produce, meat and fish, dairy and chilled, tins and dry goods, frozen, other. Cross off anything I've already said I have. Give quantities, not just item names."
This is the one that saves the actual time. A combined list grouped by aisle means I'm not backtracking through Aldi three times because I forgot the tinned tomatoes were on the same list as the pasta.
4. The leftovers and waste prompt
"Looking at this week's plan, is there anything likely to go to waste, and can you suggest one extra meal or lunch that uses up any odd amounts left over, like half a bag of spinach or two extra tomatoes?"
This single prompt has probably saved me more money than the budget constraint in prompt one. Food waste in the average UK household runs to about £60 a month according to WRAP's household food waste research, and most of that is exactly this kind of half-used bag of something nobody planned for.
A real week, with real numbers
Here's an actual week from February, not a tidied-up example. I asked for five dinners, two adults, £45 budget, using up leftover roast chicken and a bag of frozen sweetcorn. ChatGPT gave me chicken and sweetcorn chowder, a chickpea and spinach curry, fish tray bake, a stir fry using the last of the chicken, and a veggie chilli batch-cooked to double as Monday's lunch too. The combined shopping list came to 23 items, and my actual Tesco spend that week was £41.60. Not because the AI is a genius at pricing (it isn't, it has no live access to Tesco's prices and will happily guess wrong), but because a tight, specific list stops you buying the extra things you throw in on autopilot, like a second bag of salad you didn't need.
That's the bit worth being honest about. ChatGPT does not know what things cost this week at your specific shop. It will confidently tell you a shopping list is "around £35" and it might be £50. Treat the budget number as a target to build the meals around, not a receipt it can predict.
The uncomfortable bit nobody selling this trend wants to say
Here's what I've noticed after months of doing this: the prompt was never the hard part. Writing "give me five dinners under £45" takes ten seconds. The part that determines whether you eat well and stay on budget is standing in the kitchen on Wednesday night deciding whether to cook the thing on the list or order a takeaway because you're tired. ChatGPT organises information. It doesn't cook the meal, it doesn't stop you being knackered, and it doesn't know your fridge went off because the door wasn't shut.
Most people who try this quit after two or three weeks, and it's rarely because the prompts stopped working. It's because meal prep was always about willpower and time, and a chatbot can't lend you either. If you're someone who already struggles to stick to a shopping list written on paper, an AI-generated one won't magically fix that. It just makes the planning part faster, which is useful, but it's not the whole battle.
Adjusting the prompts for different goals
If the point is weight management rather than just getting dinner sorted, the prompt needs a different shape. I've written more on this in how to use ChatGPT prompts to support weight loss goals, but the short version is you need to give it a calorie or protein target, not just "healthy," because "healthy" means nothing to a model that doesn't know if you're trying to cut, maintain, or bulk.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
If money is the tighter constraint than time, I'd point you at my full breakdown of planning a week of dinners on a budget, which goes deeper into batch cooking and stretching cheaper cuts of meat across two meals.
And if you just want to see what else this tool can do beyond the practical stuff, I tested a pile of other uses in 47 fun ChatGPT prompts I've tried for personal use, some of which are just for laughs but a few crossed over into things I now use monthly.
Family version of the prompt
For a family of four with a fussy nine-year-old, the prompt needs the fussiness written in explicitly, not softened:
"Five dinners for a family of four, two adults and two kids aged 9 and 12. Budget around £60. The 9-year-old won't eat anything spicy or with visible chunks of vegetable, but will eat vegetables blended into a sauce. Keep at least three meals ones the kids will eat without a fight. Batch cook enough of one meal to freeze for a busy night later in the month."
Naming the actual fussiness rather than saying "kid-friendly" gets you meals that will work, not just meals that sound like they should.
Mistakes I made so you don't have to
- Asking for "healthy meals" with no numbers attached, which gave me suggestions with more calories than what I was already eating.
- Not telling it what was already in the freezer, so I ended up with a plan that duplicated three meals I'd already have to defrost around.
- Trusting the price estimate as gospel and being annoyed when the actual shop cost £8 more.
- Asking for the shopping list before the meal plan was finalised, which meant re-asking twice and wasting the time I was trying to save.
- Never telling it about food I dislike, only food I'm allergic to, so I kept getting mushroom-heavy suggestions I'd throw away uneaten.
A shorter version if you're in a rush
If you only take one prompt from this whole post, use this one. It combines the plan and the list in a single ask, which is less precise but faster if you're doing this at 11pm on a Sunday with one eye on the telly:
"Give me five weeknight dinners for [number] people, budget £[amount], using up [whatever's in your fridge], with a combined shopping list grouped by supermarket section at the end."
It won't be as tightly tailored as running the four prompts separately, but it takes thirty seconds and gets you 80 percent of the way there.
Related guides live in the ChatGPT Prompts and Limits: 30 Guides to Fun Prompts, Photo Edits and Daily Caps.
Try my free prompt builder before you post.
Frequently asked questions
Can ChatGPT know supermarket prices for my shopping list?
No, not reliably. It has no live connection to Tesco, Asda, or Aldi pricing, so any total it gives you is an estimate based on general knowledge, not current prices in your area. Use the budget as a target for the meals, not a promise about the final receipt.
How specific do I need to be for the prompts to work well?
Very specific. Tell it your budget, how many people you're feeding, what's already in your fridge and freezer, any dislikes (not just allergies), and how much time you have to cook on a weeknight. Vague prompts like "give me healthy meals" produce generic, unusable answers.
Will this save me money on my weekly shop?
It can, mostly because a tight, itemised list stops impulse additions, and asking it to flag likely leftovers reduces food waste, which the UK's food waste charity WRAP estimates costs the average household roughly £60 a month. It won't magically make ingredients cheaper.
Why do people give up on ChatGPT meal planning after a few weeks?
Usually not because the prompts stop working, but because the actual barrier was never the planning stage. Cooking the meal on a tired Wednesday night takes willpower and time an AI can't provide, so the tool helps most when your main problem was disorganisation rather than motivation.