Using ChatGPT to Decide What to Cook: What Works (and What Doesn't)

Using a general AI chat to decide what to cook works well for one thing: brainstorming meal ideas from whatever you have. It falls short on the parts that actually matter when you are cooking, like accurate quantities, staying with you step by step, and remembering your tastes week to week. Here is the honest breakdown of what works, what doesn't, and how to close the gaps.

What works well

A general AI chat is a strong idea machine. Type 'I have chicken thighs, rice, and half a bag of spinach, give me three dinner ideas' and you get solid, varied suggestions in seconds. It is great at riffing on a single ingredient, swapping things for an allergy, and explaining a technique you have never tried. For pure 'what could I make with this', it genuinely helps.

• Brainstorming several meal directions from a few ingredients
• Suggesting swaps when you are missing something or avoiding an ingredient
• Explaining a method, like how to get a crust on a steak or fix a broken sauce
• Drafting a rough week of meals when you give it your preferences

What doesn't work

The cracks show the moment you try to actually cook from it. Quantities and timings can be confidently wrong, because a general chat is predicting text, not testing recipes. It does not remember that you hate cilantro unless you remind it every time. And it is a wall of text, which is useless when your hands are wet and the pan is already hot.

1. Portions and cook times can be off, so you have to sanity-check them
2. It forgets your tastes and allergies between conversations unless you repeat them
3. No hands-free cooking mode, so you are scrolling a long reply mid-cook
4. No real shopping list that updates as your plan changes
5. It will happily invent a recipe that nobody has ever tested

Prompts that get better results

You can squeeze a lot more out of a general chat with tighter prompts. Be specific about constraints up front so it does not guess.

• State your hard limits first: 'no nuts, no cilantro, ready in 20 minutes'
• Give it the one thing to feature: 'build a dinner around this leftover roast chicken'
• Ask for one recommendation, not ten: 'pick the single best option and explain why'
• Ask it to list exact quantities and a numbered method, then read them back to check
• Ask for a portion note tied to your goal, like a bigger or lighter plate, without turning it into a counting exercise

Where it falls down at the stove

Even with perfect prompts, a chat window was not built for the actual cooking. You get a great-looking reply, then you are stuck holding a greasy phone, scrolling back up to find step three, while the garlic burns. There is no timer tied to the step, no voice reading it out, no way to glance and keep moving. The brainstorm is the easy 10 percent. The cooking is the 90 percent a chat does not help with.

Closing the gaps with a purpose-built tool

This is where a dedicated cooking app earns its place. The good ones keep the part a general chat does well, turning one ingredient into a meal, and add the parts it cannot: quantities scaled to your servings, a memory of your tastes and allergies so you never repeat yourself, a hands-free cook mode that reads each step aloud, and a shopping list that updates with your plan.

Pann is built exactly around this. You give it the one thing you have, it builds a real meal sized to you, remembers your dislikes and goal, and then cooks it with you step by step in voice-guided Cook Mode. You get the brainstorm and the stove help in one place, instead of a smart chat that goes quiet the moment the pan heats up.

The verdict

Use a general AI chat for inspiration and quick swaps. It is genuinely good at that. Just do not trust its quantities blindly, and do not expect it to hold your hand while you cook. For the full loop, from 'what do I make' to a finished plate, a purpose-built cooking assistant covers the gaps a chat leaves wide open.