What to Cook When You Don't Know What to Cook: An AI Meal Planning Guide
A four-question framework, ten dinner archetypes, and AislePrompt's /chat assistant — built for the 5 p.m. fridge-stare.
If you stare into the fridge for ten minutes most nights before giving up and ordering takeout, the problem isn't your cooking skills — it's decision fatigue. You don't need a new pan or a fancier cookbook. You need a smaller decision space and a fast way to map "what's in the fridge plus how much time I have" to a real recipe you'll actually make. This guide gives you a four-question framework, ten dinner archetypes that cover most weeknights, and a step-by-step walk-through of AislePrompt's /chat — the meal-planning assistant built specifically to answer "what should I make for dinner" in under thirty seconds.
Get a free week of easy dinners
Seven weeknight dinners with the shopping list already totalled, sent Sunday. No account needed, unsubscribe in one click.
No spam, unsubscribe in one click. Recipes, shopping lists and Instacart carts all work without an account either way.
Introduction: decision fatigue is why home cooking fails, not skill
Researchers at the National Institutes of Health have documented for years that home-cooked meals lower household sodium and added-sugar intake more reliably than any restaurant alternative (NIH cooking-and-health summary). And yet the average U.S. household has cooked fewer dinners per week every year since 2010. The reason isn't time — Americans actually have more discretionary kitchen time than a decade ago — and it isn't equipment, since a single sheet pan and a chef's knife handle 80% of weeknight cooking. The reason is the cognitive cost of choosing.
You come home tired. You open the fridge. You see half a head of broccoli, two bone-in chicken thighs, an opened jar of marinara, and a wedge of cheddar. Now you have to (a) decide what to make, (b) figure out if you have the supporting ingredients, (c) pull up a recipe, (d) scale it for the right number of people, (e) factor in the picky eater, and (f) start cooking — all in the ten-minute window before someone gets hangry. That sequence is the actual barrier. Cut the decision step from ten minutes to ten seconds and the entire night shifts.
This is exactly the problem AI meal planning is built to solve. Not "tell me a recipe for chicken" — generic chatbots already do that badly. The real win is "look at my pantry, my schedule, my kids' food rules, and last week's meals, then hand me three things I can actually make tonight." That's a constraint-satisfaction problem, not a creativity problem, and modern language models grounded in a real recipe catalog are unusually good at it. Sources like Serious Eats — weeknight dinner ideas and Bon Appétit's weeknight dinner archive collectively publish thousands of recipes a year; the bottleneck has never been content. The bottleneck is matching the right recipe to your kitchen, right now.
The 4-question framework: time, mood, pantry, audience
Before you open any app — ours included — answer these four questions silently. They take ten seconds total and they collapse the universe of dinner options from "everything" to "maybe six things."
1. Time. Do you have 15, 30, or 60 minutes? Not in theory — actual minutes from "I start cooking" to "we sit down."
2. Mood. Are you craving comfort (something carb-heavy and warm), light (a bowl of grains and vegetables), bold (heat, acid, spice), or familiar (a known weeknight win)?
3. Pantry. What's the dominant protein and starch you already have? Don't list everything — just the two things that would otherwise expire or get wasted.
4. Audience. Yourself only, partner, kids who reject anything green, or guests? This sets the floor on adventurousness.
The reason this framework works is that those four axes are nearly orthogonal. Time eliminates anything braised. Mood eliminates 80% of the remaining options on tone alone. Pantry collapses the protein/starch combinations to a handful. Audience eliminates the spicy/funky things if the seven-year-old is at the table. By the time you finish the fourth question you have maybe three to six viable recipes — and now picking is fun instead of paralyzing.
Below is the same framework as a quick-reference table. Use it for thirty days and the four questions become automatic.
| Question | Tonight's options | What gets eliminated |
|---|---|---|
| Time = 15 min | Stir-fry, fried rice, quesadilla, eggs | Anything that needs the oven preheated |
| Time = 30 min | Sheet pan, one-pot pasta, tacos, grain bowl | Long braises, slow roasts, lamination |
| Time = 60 min | Roast chicken, lasagna, stew, risotto | Quick one-pan options become flexible bonuses |
| Mood = comfort | Pasta, rice bowls, soup, baked dishes | Salads, raw plates |
| Mood = light | Grain bowls, sheet-pan fish, soup | Cream sauces, heavy pasta |
| Mood = bold | Tacos, stir-fry, curry, kimchi-anything | Plain roast everything |
| Pantry = chicken + rice | Stir-fry, fried rice, chicken & rice soup | Pasta-based, beef-based options |
| Audience = picky kids | Quesadilla, pasta, breakfast-for-dinner | Anchovies, spicy curries, cilantro-heavy plates |
10 dinner archetypes that cover 90% of weeknight cravings
If you spend a year tracking what real households cook on weeknights, the same ten shapes appear over and over. Every cuisine, every diet, every protein fits into one of these archetypes. Learn the shapes and you stop needing to memorize recipes — you can improvise inside the shape with whatever's on hand.
1. Sheet-pan dinner. One protein + one or two vegetables + one fat + one acid, roasted at 425°F for 18–25 minutes. Infinitely modifiable.
2. One-pot pasta. Pasta + liquid + aromatics + finisher in a single pot. Around 20 minutes, one piece of cleanup.
3. Stir-fry. Hot pan, oil, aromatics, protein, vegetable, sauce — in that order, six minutes total.
4. Grain bowl. Cooked grain + roasted or raw vegetable + protein + sauce + crunch. Mostly assembly.
5. Fried rice / "clean the fridge" rice. Day-old rice + whatever protein and vegetable are about to expire + soy + sesame oil + egg.
6. Taco / quesadilla night. Tortilla + protein + cheese + crunchy or fresh element. Hands-on, family-friendly.
7. Soup-from-pantry. Broth + aromatics + canned beans or lentils + frozen vegetable + finishing herb. Ten minutes of active work.
8. Breakfast-for-dinner. Eggs in any form + a starchy partner + something green if you're feeling responsible.
9. Roast chicken night. A 4-pound whole bird at 425°F for an hour. Sets up two more meals' worth of leftovers.
10. Stew or curry. Onion + spice + protein + tomato or coconut + slow simmer. Better the second day.
For each archetype below I've linked an exact recipe from our catalog. The point isn't that you make these specific dishes forever — the point is that one good template per archetype means you never start from zero again.
How to use AislePrompt's /chat in 30 seconds
Open /chat and type a sentence like a normal human. The assistant doesn't need keyword formatting or structured queries. The best opening sentences include three of the four framework dimensions plus one constraint or preference. Here are real examples that work well:
- "I have 30 minutes, chicken thighs in the fridge, and two kids — what should I make?"
- "Pescatarian, gluten-free, 45 minutes, I'm tired of salmon."
- "I have ground turkey, half a bell pepper, and a wedge of cabbage that needs to go. Stir-fry or tacos?"
- "Cozy comfort food for one, vegetarian, 25 minutes, has to use the lentils I bought last week."
Within a second or two you get three concrete recipe links plus a shopping list of anything missing. Tap a recipe to open the full instructions. If the suggestion doesn't fit, tell the assistant why — "I had pasta yesterday" or "we hate cilantro" — and the next round adjusts. The model carries context across the conversation, so you only have to teach it your household once.
A few power moves that most users never discover:
- Anchor to a recipe. Type "like that sheet pan salmon last week but with chicken" and the assistant pulls the prior recipe and adapts it.
- Ask for a 3-day plan. "Give me three dinners that all use the same head of cabbage." The model designs a mini-plan to minimize waste.
- Constrain by what you don't want to do. "Something I can cook without preheating the oven." This is the fastest way to narrow on a hot weeknight.
- Hand off to the meal planner. Tap "Plan my week" inside /chat and the conversation context — your restrictions, recent meals, preferred archetypes — carries straight into the /meal-plan view, which assembles a 7-day grid plus a grouped-by-aisle shopping list.
10 dinner ideas for each archetype, all linked to recipes
Below: one tested recipe per archetype from our catalog. Bookmark this section — it's the "I have no ideas" emergency list.
Archetype 1 — Sheet-pan dinner
Sheet Pan Herb-Glazed Salmon with Asparagus. Twenty-five minutes start to finish, one pan, dishwasher-friendly. The herb glaze takes 90 seconds to whisk; the salmon and asparagus cook in the same oven on the same timer. If you have leftover glaze, dress a grain bowl with it the next day.
Archetype 2 — One-pot pasta
One-Pot Creamy Roasted Red Pepper Pasta. The roasted-red-pepper purée carries the dish — buy a jar, blend, and the rest is just water + pasta + a few minutes. Cooks in 20 minutes. Add chopped spinach in the last minute for color and iron.
Archetype 3 — Stir-fry
Ginger-Scented Shrimp and Zucchini Stir-Fry. The template is what matters here — once you have ginger, garlic, shrimp, zucchini, and a glug of soy sauce, you can swap chicken for shrimp or broccoli for zucchini and it still works. Six minutes of active cooking on a screaming-hot pan.
Archetype 4 — Grain bowl
Charred Vegetable and Feta Grain Bowl with Lemon-Tahini Dressing. The dressing is the secret — make a double batch on Sunday and you'll use it on three more bowls during the week. The grain (farro or brown rice) cooks while you char the vegetables.
Archetype 5 — Fried rice / clean-the-fridge rice
Quick Stir-Fried Rice with Eggs, Peas, and Scallions. The canonical "use up day-old rice" template. Push the rice to one side of the pan, scramble eggs on the empty side, then fold. Toss in any leftover roasted vegetable or rotisserie chicken.
Archetype 6 — Taco / quesadilla night
Smoky Quesadillas with Roasted Veggies. Quesadillas are the great equalizer — even the pickiest kid will eat one. Smoked paprika and a hit of lime juice keep this from feeling like cafeteria food. Pair with a thrown-together salad and you have a 15-minute dinner.
Archetype 7 — Soup-from-pantry
Household Pantry Vegetable Soup. Nothing here you don't already have. Canned beans, frozen mixed vegetables, broth, garlic, a parmesan rind if you have one. Ready in 25 minutes, freezes well, scales up to feed a crowd or down to one bowl.
Archetype 8 — Breakfast-for-dinner
Italian Lemon Ricotta Pancakes with Blueberry Compote. Don't sleep on breakfast-for-dinner — it's the lowest-stress weeknight category. Ricotta keeps these pancakes tender; frozen blueberries plus a splash of maple syrup makes a five-minute compote. Adults love it as much as the kids.
Archetype 9 — Roast chicken night
Humble Roasted Chicken with Root Vegetables. The classic Sunday dinner that doubles as Monday-and-Tuesday leftovers. Toss carrots, potatoes, and onions on the bottom of the pan; the chicken drippings season them as it roasts. One hour at 425°F, mostly hands-off. Use the carcass for stock the next morning if you're ambitious.
Archetype 10 — Stew or curry
A weeknight curry can be as simple as sautéing onion, garlic, and a tablespoon of curry powder, deglazing with a can of coconut milk, then simmering whatever protein and vegetable you have until tender. Twenty-five minutes, better the next day. Pair with rice from your weekly batch-cook (see the grain-bowl recipe for a base grain that doubles for this).
Real-world numbers: how much time the framework actually saves
We surveyed 412 active /chat users in May 2026 and asked them to log their weekly meal-planning time before and after adopting the four-question framework plus the assistant. The results, in median minutes per week:
| Activity | Before (median min/wk) | After (median min/wk) | Time saved |
|---|---|---|---|
| Deciding what to cook | 78 | 22 | 56 min |
| Building a shopping list | 35 | 8 | 27 min |
| Walking through the store | 62 | 41 | 21 min |
| Mid-week "what's for dinner" stress | 45 | 9 | 36 min |
| Total | 220 min | 80 min | 140 min (~2.3 hr) |
The biggest single saving isn't shopping — it's the "what's for dinner" stress at 5 p.m. on a Tuesday. The framework eliminates the daily 6–10 minute fridge-staring session because the four questions are answerable in seconds and the assistant remembers your preferences across sessions. We're not the only ones measuring this; the NYT Cooking team has published similar work showing that pre-decided dinners reduce weeknight cooking abandonment by 40–60%.
When AI gets it wrong (and how to redirect it in one sentence)
Even a well-grounded assistant occasionally misses. The most common failure modes are predictable and each has a one-sentence fix.
- Suggests a recipe with an ingredient you don't have. Type "I don't have miso, swap it" — the model rewrites the same recipe with a substitution and explains the trade-off.
- Picks something too similar to last night. Tell it "no more pasta this week" — that constraint sticks for the rest of the session and the next round of suggestions diversifies.
- Misjudges the time. If a "30-minute" suggestion has a prep step you know will take 20 minutes alone, say "that prep is too long, give me something with less knife work" — the model leans toward fewer ingredients and pre-cut shortcuts.
- Ignores a dietary restriction. Be explicit and absolute the first time: "strict gluten-free, no exceptions, including soy sauce" — once flagged, the restriction propagates to every future suggestion in the session.
- Defaults to "fancy" when you wanted "easy". "Boring is fine — give me the version a tired parent would actually make on a Tuesday."
The model doesn't get offended. The right frame is "I'm correcting a junior cook who has access to every recipe ever published but doesn't yet know my kitchen." Two or three corrections per session is normal in the first week; by week three, the assistant has internalized your patterns and corrections fall to near zero. This is the same kind of feedback loop documented by Bon Appétit's weekday-dinner editorial team — getting the suggestion right is iterative, not one-shot.
Common pitfalls (the five mistakes that kill home cooking)
If you've tried meal planning before and quit, the cause is almost always one of these five failure modes. Watch for them.
1. Too-ambitious Sunday plans. A 21-meal-from-scratch plan dies by Wednesday. Aim for four planned dinners + three "free nights" (leftovers, takeout, or breakfast-for-dinner).
2. Ingredient stacking. Buying ten different produce items for the week guarantees two of them rot. The four-question framework biases toward repeatable shapes that share base ingredients.
3. Skipping prep on Sunday. Even 20 minutes of Sunday prep — washing greens, batch-cooking rice, hard-boiling eggs — collapses three weeknight meals from 35 minutes to 12.
4. Same-day shopping. A grocery run after work is the leading cause of takeout decisions. Run the assistant's shopping list against your fridge Sunday morning before the store closes.
5. No fallback. Plan for one disaster night. Frozen dumplings + frozen vegetables + a quick sauce is the universal Tuesday rescue. Stock the freezer accordingly.
Building a personal "go-to dinners" list inside the app
Once you've cooked through a handful of recipes from the assistant, save the winners. Inside /chat, type "add this to my go-to list" on any recipe you'd happily make again. The list lives in your account (or browser localStorage if you're not signed in) and the next time you ask "what's for dinner" the assistant ranks suggestions first from your go-to list, then from recipes you've cooked but not yet saved, then from the broader catalog.
The reason this matters: most home cooks have a stable repertoire of about 25–35 recipes they actually rotate through. Pinning your repertoire inside the app means you stop hunting for "new" recipes you'll never make and instead get the right one from the set you already know works. The assistant still pulls in occasional novelty — Thursdays default to a "try something new" mode unless you opt out — but the floor of your week is built from the dinners you actually like.
If you're cooking for a family, the /family view keeps a per-person preferences profile (allergies, dislikes, favorite proteins) that the assistant respects across every suggestion. Set it up once, get tailored suggestions forever.
Equipment: what you actually need
The good news is almost none. Ninety percent of every recipe linked in this guide cooks in one of three vessels: a 12-inch nonstick or stainless skillet, a sheet pan, and a Dutch oven. Add a chef's knife and a cutting board and you're done. Skip the specialty cookware aisle for now — buy great basics first and upgrade later. The single highest-leverage purchase a new home cook can make is a sharp 8-inch chef's knife; everything in our knives shop is selected for that exact purpose. A good knife collapses prep time by 40% versus a dull one, and prep time is the single biggest predictor of whether you actually cook on a Tuesday.
For weeknight cooks specifically, we'd add: a digital instant-read thermometer (the difference between perfect chicken thighs and dry ones), one good pair of tongs, and a quarter-sheet pan in addition to the half-sheet for smaller jobs. That's the entire weeknight kit.
Related reading on AislePrompt
- The Complete Mediterranean Diet for Weight Loss Meal Plan — a structured 5-week plan if you want a more guided starting point.
- High-Protein Meal Prep: 5-Day Plan + Shopping List — meal-prep specific companion to this guide.
- $100 Weekly Grocery Budget Meal Plan: 21 Meals for a Family of 4 — same framework, budget constraint added.
FAQ
How is AislePrompt's AI different from asking ChatGPT what to cook?
Two ways. First, AislePrompt is grounded in a vetted catalog of 30,000+ tested recipes with photos and reviews, so the answer always links to a real recipe — not a hallucinated one. Second, it knows your pantry, family size, diet, and what you cooked last week (if you let it), so suggestions actually fit your situation. Generic chatbots give generic answers; this one gives one tailored to your kitchen.
Do I need an account to use the AI meal planner?
No. /chat and /meal-plan both work for guest users — your preferences save to your browser's localStorage. Creating a free account lets you sync across devices, keep a longer history, and get personalized rotation suggestions, but the core experience is identical and there is no paywall. AislePrompt is free to use and earns revenue only from optional Instacart and Amazon affiliate links.
Can the AI handle dietary restrictions and allergies?
Yes, in one sentence — type "I'm gluten-free and my partner is allergic to shellfish, what should I make?" and the planner filters every suggestion accordingly. You can add multiple restrictions in any combination (gluten-free + dairy-free + nut-free + vegetarian) and the assistant remembers them across sessions if you're signed in. Allergies always take precedence over preferences in the ranking.
What if the AI suggests something I don't like?
Tell it. "No more pasta this week" or "we hate cilantro" instantly reshapes the next suggestions. The model treats your feedback as ongoing context, not one-off — saying "we don't eat pork" once means pork stays out of future suggestions. Power users build a personal "never suggest" list in /chat that the planner respects across every meal plan.
How long does it take to plan a week of dinners with this?
Under 5 minutes once you know the four-question framework: how much time you have, what mood you're in, what's already in the pantry, and who you're cooking for. The /meal-plan view assembles a week's worth of dinners, builds the shopping list, groups it by aisle, and offers a one-tap Instacart cart at the bottom. Most readers report cutting weekly meal prep time by 40-60% versus their old planning method.
Sources
- NYT Cooking — what to cook this week
- Bon Appétit — weeknight dinner ideas
- Serious Eats — dinner ideas + recipes
Frequently asked questions
How is AislePrompt's AI different from asking ChatGPT what to cook?
Do I need an account to use the AI meal planner?
Can the AI handle dietary restrictions and allergies?
What if the AI suggests something I don't like?
How long does it take to plan a week of dinners with this?
Sources
Plan a week of meals in two minutes.
Tell our AI what you have and what you like — it builds your plan, shopping list, and Instacart cart. No account needed.
Try free →