TL;DR
Paste your fridge contents, dietary limits, and a per-meal time budget into ChatGPT, Claude, or Gemini and ask for a 7-day grid plus one aisle-grouped shopping list. Use the copy-ready prompt below. For a hands-off app instead, Mealime (free, Pro $5.99/mo) or Samsung Food+ (~$60/yr, scans your fridge from a phone photo) do the same job with a built-in grocery list. The trick that actually saves money: force ingredient reuse across 2-3 meals and route every dinner into a next-day lunch, which cuts the week from 21 cooks to roughly 14.
The task
You open the fridge twice a day, stare, and order takeaway. A week of meal planning would kill the decision fatigue, but a generic plan from a magazine assumes unlimited ingredients and unlimited time. You want a plan that starts from what you already have, fits the actual time you have to cook on a Tuesday, and gives you one consolidated grocery list, not 21 separate ones.
This is a near-ideal job for a general AI chatbot: it is pattern-matching ingredients into recipes, sequencing leftovers, and deduplicating a list. As of June 2026, GPT-5.5 (the default model in ChatGPT), Claude Sonnet 4.6, and Gemini 3.1 Pro all read an uploaded photo of your fridge shelf and can hold a full week’s context in one chat, so you can refine the plan over several messages instead of starting over.
When AI helps, and when it does not
AI is good at combining your ingredient list into recipes, sequencing meals so leftovers reduce prep time, and writing a deduplicated shopping list. It is poor at knowing whether a recipe tastes good to you: flag the meals you want to repeat after week 1 so the next plan calibrates. It is also a poor source for medical or allergy-specific guidance. List restrictions explicitly in the prompt, then cross-check ingredients on every recipe, and consult a registered dietitian when the stakes are clinical (a diagnosed allergy, pregnancy, a managed condition).
One concrete failure to expect: AI will confidently put a “15-minute beef bourguignon” or a braise that actually needs 90 minutes into a weeknight slot. Treat every time-to-table number as a claim to verify, not a fact.
Chatbot vs dedicated app: which to use
If you want full control and a free option, a general chatbot is fine. If you want fewer steps every week, a meal-planning app keeps your preferences, builds the list, and (on paid tiers) reads your fridge from a photo. Prices are as of June 2026.
| Tool | Free tier | Paid | Best for |
|---|---|---|---|
| ChatGPT (GPT-5.5) | Yes (ad-supported US Free, tight limits) | Plus $20/mo | Most flexible; reads fridge photos; full-week context in one chat |
| Claude (Sonnet 4.6) | Yes (limited) | Pro $20/mo | Clean tables and lists; strong at the consolidated shopping list |
| Gemini 3.1 Pro | Yes | Google AI Pro $19.99/mo | Tight Google Workspace / Keep users |
| Mealime | Yes, genuinely usable | Pro $5.99/mo, $49.99/yr | Set-and-forget weeknight dinners (recipes ≤30 min) |
| Eat This Much | Limited | ~$5/mo annual | Hitting exact calorie / macro targets |
| Samsung Food | Yes, ad-supported | Food+ ~$60/yr ($4.99/mo) | Vision AI fridge-photo scanning (40,000+ ingredients), 7-day plans |
Note on Samsung Food: the free tier gives 3-day recommendations; Food+ unlocks 7-day plans, ad removal, pantry management, and the Vision AI camera scan. A general chatbot already does the photo scan free if you take the picture yourself, so the app’s value is the saved weekly clicks, not the recognition itself.
What to feed the AI
- Current fridge / pantry contents (rough is fine; or just upload a photo of the shelf)
- Dietary preferences and hard restrictions (vegetarian, no peanuts, halal, kosher, diabetic)
- Realistic time per meal: breakfast ≤10 min, lunch ≤20 min, dinner ≤35 min
- How many people, including kids’ portions and any packed lunches
- Budget ceiling for the shopping list, if any
- Cooking skill level, honestly; AI will overestimate it
Copy-ready prompt
Plan 7 days of meals.
Fridge / pantry: [list, or "see attached photo"]
Diet and restrictions: [list]
Time budget per meal: [breakfast / lunch / dinner]
People: [number, with ages]
Budget ceiling: [amount or "flexible"]
Skill level: [basic / intermediate / advanced]
Return:
1. A 7-day table: breakfast / lunch / dinner per day
2. Each meal: name, key ingredients, time-to-table in minutes, prep notes
3. A "leftovers strategy": which dinners become next-day lunches
4. One consolidated shopping list, grouped by aisle (produce / protein / pantry / dairy / frozen)
5. A "swap menu": 2 alternates per day for when I am not in the mood
6. Total estimated grocery cost, if I gave a budget
Prefer reusing ingredients across 2-3 meals. Avoid recipes that need 5+ unique ingredients per meal.Tight-schedule variant: “Same plan, but all dinners must be 20 minutes or less, and 3 of them must be one-pan.”
Budget variant: “Re-do the plan to land under [amount]. Show the running total per line on the shopping list.”
How to check the output is usable
- The fridge contents you listed are actually used in at least 3 meals
- No meal needs more than 5 ingredients (after deduplication across the week)
- The time-to-table claims match reality; if AI claims 15-minute beef bourguignon, push back
- The shopping list is grouped by aisle, so walking the store is linear, not a zigzag
- At least one dinner becomes lunch, so leftovers shrink the cook count from 21 toward ~14
Common mistakes
- Five unique ingredients per meal. The shopping bag explodes and half of it rots. Force reuse in the prompt.
- No leftover strategy. You cook 21 meals from scratch and give up by Wednesday.
- Letting AI invent measurements. Models reproduce ratios from training data unevenly; for baking and anything precise, verify quantities against a known recipe source before you shop.
- Skipping the swap menu. One bad-mood day with no alternate breaks the whole week.
- Forgetting packed lunches in the headcount. You cook 2 portions and need 4.
FAQ
- Is a paid AI subscription worth it just for meal planning? Usually not on its own. The free tiers of ChatGPT, Claude, and Gemini all handle a weekly plan. If you want zero weekly effort, a $5-6/mo app like Mealime or Samsung Food+ is cheaper than a Plus subscription and keeps your preferences between weeks.
- Can I just photograph my fridge? Yes. GPT-5.5, Claude Sonnet 4.6, and Gemini 3.1 Pro all read an uploaded shelf photo and list the ingredients. Add a sentence about what is in the freezer and pantry, since the photo only shows the front row.
- What about allergies? Be explicit in the prompt, and cross-check ingredients on every recipe. AI is not a medical source; for a diagnosed allergy, verify with a dietitian.
- How do I handle a sudden change of plans? Use the swap menu you asked for. Re-plan the whole week only on weekends so leftovers still chain correctly.
- Can AI estimate calories? Roughly. Treat the numbers as ±20%. For exact macro tracking, Eat This Much or a logging app is built for it.
Related
- Weekly planning — same logic for the rest of your week
- Decluttering plan — energy-saver pattern for the home
- Habit tracker AI — turn “I plan meals” into a habit
- Household task schedule — chore split that includes cooking
- Task prioritisation AI — when “cook dinner” loses to “answer Slack”
- Project Planning Prompts: From Goal to Sprint Plan