An AI model can plan a week of meals in under a minute if you tell it four things: who’s eating, what they won’t or can’t eat, how much time you have on weeknights, and roughly what you want to spend. Then ask it to turn the plan into a grocery list grouped by store section. You don’t need a meal-planning subscription for this, and if you use a model that runs on your phone, your diet and health details stay with you.
The quality of the plan depends almost entirely on what you tell it, so start there.
What to tell the AI before it plans #
A good meal plan starts with a short profile. Write it once and reuse it.
- Who’s eating: “Two adults and a 7-year-old who won’t eat anything spicy.”
- Restrictions: allergies, intolerances, vegetarian days, foods you dislike.
- Time: “Weeknights 30 minutes max, one longer cook on Sunday.”
- Budget: a rough weekly amount, or “keep it cheap.”
- Equipment: air fryer, slow cooker, no oven, one pot only.
- What’s already in the kitchen: leftovers, a bag of rice, half a cabbage.
- Goals: more vegetables, more protein, fewer takeout nights, batch cooking for lunches.
A meal-plan prompt you can copy #
Plan 5 weeknight dinners and 1 weekend meal for two adults and a 7-year-old. No spicy food, no mushrooms, one vegetarian night. Weeknights must take 30 minutes or less. Budget is tight. Reuse ingredients across meals to cut waste. For each meal give the name, a one-line description and the main ingredients. Don’t include full recipes yet.
Then follow up:
- “Swap Wednesday for something using the leftover chicken from Monday.”
- “Give me the full recipe for Thursday.”
- “Make Sunday’s meal big enough for two lunches.”
- “Now write the grocery list, grouped by produce, dairy, meat, pantry and frozen. Leave out rice and olive oil, I have them.”
Asking for the outline first, then details, works much better with smaller models than asking for everything in one reply.
How to turn the plan into a grocery list #
- Tell it what you already have so it doesn’t list it.
- Ask for quantities for the whole week, combined across meals (“3 onions total,” not three separate lines).
- Group by store section so you walk the aisles once.
- Ask for cheaper swaps where they don’t matter: frozen spinach for fresh, thighs for breasts.
- Copy the list into your notes app, or keep it in the chat.
Check the quantities against the recipes. AI is good at structure and weaker at arithmetic, so it can double-count an ingredient or miss one. Can local AI do math explains why and how to reduce errors.
Using a photo of your fridge #
If your AI app understands images, take a photo of the fridge or pantry and ask, “What could I make tonight with what’s here?” It’s a quick way to use up leftovers. Vision models can misidentify items in a crowded fridge, so correct it if needed (“that’s feta, not tofu”).
What an AI meal planner gets wrong #
- Nutrition numbers are estimates. Calorie and protein counts from a chatbot are rough guesses, not lab values. Use nutrition labels or a nutrition database if the numbers matter to you.
- Allergies need human checking. An AI can suggest a dish that contains a hidden allergen, like fish sauce or nut oils, or forget a restriction halfway through a long chat. Read every recipe with the allergy in mind.
- Medical diets need a professional. For diabetes, kidney disease, eating disorders or pregnancy, a registered dietitian or your doctor should set the plan. AI can help you cook within it.
- Prices are guesses. A local model doesn’t know what things cost at your store this week.
- It forgets. In a long conversation, small models can lose track of earlier constraints. Restate the key ones, or keep them in a system prompt.
Why keep meal planning private? #
A meal plan says more about you than it looks. Diet restrictions can reveal health conditions, religious practice, pregnancy, weight goals or a history of disordered eating. Cloud AI services store what you type, and meal-planning apps often build profiles to recommend products.
A local AI model avoids that. Personal LLM runs open models like Qwen 3.5 and Gemma 4 on your iPhone or Android phone, with no account and nothing sent to a server. A few features suit meal planning in particular:
- A system prompt per chat. Put your household profile in a “Kitchen” chat’s system prompt so every plan follows it without retyping.
- Photos. With image support added, ask about a fridge shelf, a recipe card or a label.
- Offline. It works in the basement of a grocery store with no signal.
- Search and pin. Pin the chat with your family’s favorite meals so it’s easy to find.
Small models give simpler plans than the big cloud chatbots, but meal planning mostly depends on what you tell them, which is where they do well. Our system prompt examples include more profiles you can adapt.
A sample weekly workflow #
- Sunday, 5 minutes: ask for next week’s dinners using your saved profile and what’s left in the fridge.
- Adjust: swap meals you don’t like, ask for one batch-cook.
- List: generate the grouped grocery list and copy it.
- Shop: check off items in your notes app.
- Cook: ask for the full recipe on the night, with timings for your equipment.
- Next week: tell it what the family liked and what nobody ate.
If you’re pairing this with training, the same approach works for exercise; see a private AI workout plan.
Frequently asked questions #
Can AI make a meal plan for free? #
Yes. Free tiers of cloud chatbots can, and so can free apps that run a model on your phone. You don’t need a paid meal-planning app to get a weekly plan and grocery list.
Is an AI meal plan healthy? #
It can be balanced if you ask for that, but it isn’t medical or dietary advice. Nutrition numbers are estimates. For medical diets, work with a dietitian or doctor and use AI to help with recipes and lists.
How do I make AI remember my family’s preferences? #
Cloud chatbots may store a memory of you on their servers. With a local model, put your preferences in a system prompt for your meal-planning chat, so they apply every time without being stored anywhere else.
Can AI plan meals around allergies? #
It can try, and it’s useful for ideas, but it can miss hidden ingredients or forget a restriction. Always check recipes and labels yourself when an allergy is involved.
Can AI tell me what to cook with what’s in my fridge? #
Yes. List the ingredients or, with a vision-capable app, take a photo and ask for dinner ideas. Correct anything it misidentifies before you start cooking.