How to write a useful AI recipe request

You do not need prompt-engineering jargon. You need to describe the kitchen reality that matters.

Recipe generators are most useful when the request contains enough constraints to avoid obvious mismatches. A long message is not automatically better. Five concrete details usually do more than a paragraph of vague preferences.

1. Say who is eating

Give the number of servings and any dietary requirements. Treat allergies separately: generated output may make mistakes, so verify every ingredient label and cross-contamination warning yourself.

2. Give a real limit

“Cheap” means different things to different people. “Under €8 total” is testable. The app’s prices are estimates, not shop guarantees, so leave a margin.

3. Name the ingredients that matter

List must-use items and explicit exclusions. You do not have to inventory the whole kitchen. “Use the open can of tomatoes; no mushrooms” is enough to shape the result.

4. Include the time and equipment

Twenty minutes, one pan and no oven describe a different meal from an open-ended weekend dinner. Mention what you cannot use.

5. Describe the outcome

Words such as crispy, comforting, light, spicy or lunch-box friendly provide direction without forcing a specific dish.

Too broadMore useful
Give me dinner.A comforting vegetarian dinner for two under €7, ready in 25 minutes.
Use my vegetables.Use one zucchini and half a pepper; avoid mushrooms and keep it one-pan.
Healthy pasta.A high-fibre pasta for three with beans, lots of vegetables and no cream.

Then refine one thing at a time

If the first result is close, request a clear change: “Make it vegan but stay under the same budget.” If you want a genuinely new idea, ask for a completely different alternative with the same constraints.

Remember

AI output can contain incorrect quantities, unsafe cooking advice or unsuitable ingredients. Check food safety, allergens and medical dietary requirements independently.