Your First Culinary AI Prompts
The single most common reason chefs give up on AI is that they treat it like Google. They type "spring menu ideas" and get back ten generic asparagus dishes that read like a Pinterest board from 2014.
That is a prompting problem, not an AI problem. This lesson teaches you a simple framework that turns AI from a search engine into a sous-chef who already knows your kitchen.
What You'll Learn
- The CRISP prompt framework for chefs
- How to give AI enough context to actually help you
- Three copy-paste prompts you can use today (recipe ideation, menu copy, allergen swap)
- How to iterate on a weak first answer instead of giving up
Why "Spring Menu Ideas" Doesn't Work
When you type "spring menu ideas" into ChatGPT, you are giving it almost no context. It has no idea what kind of restaurant you run, who eats there, what your average check is, what equipment you have, what your team can execute on a Friday night, or what season the produce in your region is actually in. So it gives you the median answer — which is the median of every food blog ever written about spring menus. Asparagus. Peas. Lemon. Generic.
The fix is to give AI a richer brief.
The CRISP Prompt Framework
Use this five-part structure for any non-trivial prompt:
- C — Context. Who you are, what your kitchen is, who eats there.
- R — Role. What hat you want the AI to wear ("act as my menu R&D chef").
- I — Instruction. The specific task.
- S — Specifics. Constraints — budget, equipment, dietary requirements, time, season.
- P — Pattern. The format you want back (table, bullet list, recipe card, plain prose).
Here's the contrast.
Weak prompt:
Give me spring menu ideas
CRISP prompt:
Context: I am the chef de cuisine at a 60-seat New American bistro in
Portland, Oregon. Average check $58. Our regulars are educated 30-55 year
olds who care about local sourcing.
Role: Act as my menu R&D collaborator.
Instruction: Give me 8 candidate dishes for a spring tasting menu (May-June).
Specifics:
- Use Pacific Northwest ingredients in season in May (morels, spring onions,
fava beans, halibut, asparagus, rhubarb, strawberries, fresh peas)
- Mix of 3 vegetable-forward, 3 seafood, 2 land protein
- Food cost target: 28%
- We have a combi oven, induction tops, no sous vide circulators
- 2-cook line, so nothing with more than 4 plate components
Pattern: For each dish give me:
1. Dish name (one line)
2. 2-sentence concept
3. Hero ingredient
4. Why it works for the season and the audience
The CRISP prompt gives you the answer of a sous-chef who has worked at your restaurant for two years. The weak prompt gives you the answer of a stranger.
Three Prompts You Can Use Today
1. Recipe ideation around a seasonal ingredient
Act as my menu R&D chef.
Hero ingredient: black garlic (we just got 5 lb in from a local fermenter,
deep umami, slight sweetness)
Restaurant: 80-seat modern American steakhouse, Chicago, average check $95.
Give me 6 dish concepts that feature black garlic without making it the
star -- I want it as a supporting flavor in dishes that read familiar to a
steakhouse guest. Mix of starter, side, sauce/condiment, dessert.
For each: 1-line concept, hero protein/element, why a steakhouse guest
would order it.
2. Menu copy from a scribbled dish concept
Act as my menu writer.
House voice: confident, specific, no clichés. Banned words: artisanal,
elevated, mouthwatering, decadent, kissed, journey, symphony, whisper.
Dish: grilled half quail, cornbread puree, pickled chanterelles, sorghum
glaze, watercress.
Write three menu description options:
- 6 words
- 12 words
- 20 words
Use only ingredients I listed.
3. Allergen substitution draft
Act as my allergen-aware sous chef.
Original dish: pan-seared scallops, cauliflower puree (with butter and
cream), brown butter, hazelnut crumble, sherry reduction.
Guest constraints: dairy-free, tree-nut-free.
Give me a substitution plan that keeps the texture and flavor architecture
of the original. List every ingredient swap with a 1-sentence reason.
Important: I will verify every swap against my actual product labels and
cross-contamination protocol before serving.
That last line — the verification reminder — is one of the most important habits to build. Always tell AI you will verify allergen output. It sets the right mindset and reminds you to do it.
When the First Answer Is Weak
You will get weak answers. The fix is almost always one of these:
- Be more specific about what you don't want. "No dishes built around a puree" cuts 40% of typical AI suggestions.
- Give an example of what you do want. "Here's a dish I love: [paste]. Now give me 5 more in that voice."
- Ask it to critique its own work. "Now rank these 8 dishes from most to least likely to sell in our market. Be honest."
- Switch tools. If ChatGPT is repeating itself, paste the same prompt into Claude. The voices are different and often one will click.
Don't accept the first weak answer and walk away. Iterating is the whole skill.
Key Takeaways
- AI's biggest weakness for chefs is missing context — fix it with the CRISP framework
- CRISP = Context, Role, Instruction, Specifics, Pattern
- Three reusable prompts get you started: recipe ideation, menu copy, allergen swap
- Always include a "verify against real labels" reminder on allergen work
- Iterate on weak answers — be specific about what you don't want, give examples, or switch tools