Guides · Updated 2026-10-24

AI for Restaurant Menus: Descriptions, Pricing, Specials and Allergens (Copy-Paste Prompts)

Your menu is the one piece of marketing every customer reads. Free AI tools such as ChatGPT, Google Gemini and Microsoft Copilot can help you write clearer, more appetising descriptions, plan specials around stock you need to move, check food cost before you set a price, and adapt the same menu for delivery apps, social posts and other languages. This guide gives you copy-paste prompts for each job, a simple way to review your menu with AI, and the allergen rules AI must never be trusted with.

Short answer: give the AI your dish names, the real ingredients and your prices, and ask it to rewrite each description in 10–15 words using only those ingredients. Use AI to draft descriptions, specials, delivery-app versions and translations, and to do the food-cost maths you've given it numbers for. Never let it decide allergen information, invent ingredients or origins ("locally sourced", "homemade") or set prices without your real costs. A person who knows the recipe checks everything before it's printed.

Replace anything in [brackets] with your details. The prompts work in the free versions of the main AI assistants.

What AI is good at on a menu, and what it isn't

1. Write menu descriptions that sell (honestly)

Good descriptions tell the customer what's on the plate and why they'd want it, in a few words. The safest way to use AI is to give it the exact ingredients and forbid it from adding anything.

The master prompt:

"You are helping a [type of restaurant, e.g. family Italian / neighbourhood café / burger takeaway] write its menu. Our tone is [warm and simple / playful / refined]. Rewrite each dish below with a description of 10–15 words. Use only the ingredients and cooking methods I list. Don't invent ingredients, origins, awards or health claims, and don't use the words 'delicious', 'mouth-watering' or 'world-famous'. Keep the dish names as I wrote them. [Paste: dish name, ingredients, cooking method, price]"

Before and after (example):

• Before: Chicken burger: chicken, bun, lettuce, mayo, chips. [price]
• After: Buttermilk fried chicken thigh, crisp lettuce and garlic mayo on a toasted bun, with chips. (Only true if your chicken really is buttermilk-fried, it's thigh meat and the mayo has garlic. Fix the draft to match your kitchen.)

More description prompts:

1. "Give me 3 versions of each description: short (under 8 words), standard (10–15 words) and descriptive (20–25 words). Same ingredients only."
2. "Rewrite these descriptions so they're easier to read aloud by a waiter in under 5 seconds each."
3. "Point out any description that's vague (e.g. 'chef's special sauce') and suggest a clearer version, based only on the ingredients I list for that sauce: [ingredients]."
4. "Write a one-line intro for each menu section ([starters, mains, sides, desserts, drinks]) that fits our tone. Max 12 words each."
5. "Suggest clear labels for these dishes, such as 'vegetarian' or 'vegan', based only on the ingredients listed. Flag any dish where you're unsure, and don't label anything 'gluten-free'."

On that last prompt: treat AI's labels as a starting list for you to check, not an answer. A dish can look vegetarian from the ingredients you typed and still use a stock or sauce that isn't.

2. Plan specials and seasonal dishes

Specials are where AI saves the most thinking time, especially when you build them around what you already have in the cool room.

6. "I have these ingredients to use up this week: [list with rough quantities]. Suggest 5 specials a [type of restaurant] could make with mostly these ingredients and our existing kitchen equipment ([list]). For each, give a name, a 12-word description and the extra ingredients needed."
7. "Write a weekly specials board for [day range] with one starter, two mains and one dessert, using these dishes: [list]. Include short descriptions and leave prices as [price]."
8. "Suggest 10 names for a [dish] that fit our [tone] menu. Avoid puns that are hard to say out loud."
9. "Plan a [season] menu change for a [type of restaurant]: suggest which 4 dishes to swap for seasonal versions, using produce that's in season in [your country/region]. Mark anything you're unsure about so I can check with my supplier."
10. "Write a set menu for [occasion, e.g. Mother's Day / a work function] at [price] per head with 3 choices per course from these dishes: [list]. Add one line I can use on social media to promote it."

3. Check food cost and price with AI

AI is useful for menu maths as long as you give it the numbers and check the working. Never accept a price it suggests without your real recipe quantities and supplier prices.

11. "Calculate the food cost of this dish per portion. Ingredients and quantities per portion: [list]. Supplier prices and pack sizes: [list]. Show each line of the working. Then tell me the food cost percentage at a menu price of [price] [including/excluding tax], and what the price would need to be for a [target]% food cost."
12. "Here are 10 dishes with food cost per portion and current price: [list]. Rank them by gross profit per portion (price minus food cost, excluding tax) and flag any with a food cost over [target]%."
13. "Our supplier has raised [ingredient] from [old price] to [new price] per [unit]. Which of these dishes use it and how much does each portion's cost go up? [Paste recipes]"
14. "Suggest 3 ways to bring this dish's food cost down to [target]% without making the portion look smaller: [recipe, quantities and costs]. Don't suggest swapping the main protein."

Watch for: tax. Make it clear whether your menu prices include sales tax or VAT, because food cost percentage is usually worked out on the price excluding tax. If the AI's numbers look off, ask it to "show the calculation again line by line". It's easy to spot a slipped decimal that way.

4. Review your menu with your sales data

A common way to review a menu is to sort each dish by two things: how popular it is (how many you sell) and how profitable it is (gross profit per portion). That gives four groups: popular and profitable dishes to keep and feature, popular but low-margin dishes to re-cost, profitable but slow sellers to promote or describe better, and slow, low-margin dishes to rethink or remove. AI can do the sorting for you if you paste in numbers from your till or POS system.

15. "Here is last month's sales for each dish: name, number sold, menu price excluding tax, food cost per portion. [Paste table]. Calculate gross profit per portion and total gross profit for each dish. Then sort them into four groups: high popularity and high profit, high popularity and low profit, low popularity and high profit, low popularity and low profit, using the average number sold and the average gross profit per portion as the dividing lines. Present it as a table and suggest one action for each dish."
16. "For the dishes in the 'low popularity, high profit' group, rewrite their descriptions to be more appealing, using only these ingredients: [list]. Also suggest where on the menu they could go to be noticed more."
17. "Our menu has [number] dishes. Suggest which [number] to cut to make the menu easier to choose from and the kitchen faster, based on this sales and prep-time data: [paste]. Explain each suggestion in one line."

Paste dish-level totals only. There's no need to share customer names, card details or staff information to do this.

5. Adapt the menu for delivery apps, QR menus and social posts

The same dish usually needs a slightly different description on a delivery app (photo-first, short, clear about portion size) than on your printed menu.

18. "Rewrite these dishes for a delivery app listing. For each: a clear dish name, a description under 100 characters that says what's in it and the portion size ([size]), and a note if it travels badly (e.g. fries, ice cream). [Paste dishes with ingredients]"
19. "Group these dishes into delivery-app categories that make sense for someone ordering on their phone, and suggest 3 combo deals using only these dishes and prices: [list]. Show the saving for each combo."
20. "Write a QR-code menu version of our menu with short descriptions, clear section headings and a line at the top: 'Please tell your server about any allergies before you order.'"
21. "Write 5 social media captions announcing our new [dish/menu], under 40 words each, with one call to action ([book / order / visit]). Don't make health claims."
22. "Write photo shot notes for our new menu: for each of these 6 dishes, suggest the angle, plate and background that would make it look best on [Instagram/delivery app], shot on a phone."

Delivery platforms set their own rules for menu content, prices and fees, and they change. Check the platform's own merchant help pages before you make big changes.

6. Translate your menu (with a human check)

AI translation is good enough for a first draft and fast for multilingual areas or tourist trade, but food names are where translations go wrong.

23. "Translate this menu into [language]. Keep dish names that are well known in their original language (e.g. 'risotto', 'bobotie', 'pad thai') and add a short translated description underneath. List any term you weren't sure about at the end."
24. "Check this [language] menu translation for words that sound odd, overly formal or wrong for a restaurant menu: [paste]."

Have a fluent speaker, ideally someone who knows food, read it before you print it. Allergen information in a translated menu must still match your English source exactly.

Allergens: the rules AI can't follow for you

This is the part of a menu where an AI mistake can hurt someone. Never ask an AI tool to work out which allergens a dish contains, and never copy allergen information from an AI draft. AI only knows the ingredients you typed, not the stock, sauces, oils, garnishes, cooking oil shared with other foods, or a supplier's product that changed last week.

What the law requires depends on where you are. Two examples:

• UK (England, Wales and Northern Ireland): food businesses must give customers allergen information for non-prepacked food, such as restaurant meals, for any of 14 allergens: celery, cereals containing gluten, crustaceans, eggs, fish, lupin, milk, molluscs, mustard, tree nuts, peanuts, sesame, soya, and sulphur dioxide/sulphites above a set level. See the Food Standards Agency's allergen guidance for food businesses.
• United States: federal law names nine major food allergens: milk, eggs, fish, crustacean shellfish, tree nuts, peanuts, wheat, soybeans and sesame (sesame was added from 1 January 2023). Federal labelling rules mostly cover packaged food, and state and local rules for restaurants vary. See the FDA's page on food allergies.

Elsewhere, check your national or local food safety authority. Whatever the rules, a safe process looks like this:

• Allergen information comes from the recipe and supplier specifications, written and checked by a person who knows the dish.
• Keep it in one place (an allergen matrix or recipe sheet) and update it the day a recipe or supplier product changes.
• Use AI only for formatting, for example: "Turn this allergen table I've written into a neat one-page staff sheet. Don't add, remove or change any allergen." Then compare the output with your original line by line.
• Put a clear note on every menu version, printed, QR and delivery, asking customers to tell staff about allergies before ordering.
• Never claim "allergen-free" or "gluten-free" unless you can back it up and meet the rules where you are.

Other menu claims to keep honest

• Origins and sourcing: only say "local", "free-range", "organic", "grass-fed" or a specific farm or region if it's true and you could show proof.
• "Homemade" and "fresh": if a sauce or dessert is bought in, don't let the AI call it homemade.
• Health and nutrition: avoid claims like "healthy", "low-fat" or "superfood" unless you've checked what's allowed where you trade. Some places also require calorie information on menus for larger chains, so check if that applies to you.
• Photos: AI-generated food images can look nothing like your real dish. Use photos of your actual food on menus and delivery apps.

A 30-minute menu refresh with AI

1. Gather your current menu, each dish's real ingredients and prices, and last month's sales per dish if you have them.
2. Run the master prompt (section 1) on one menu section at a time. Edit each description so it matches your kitchen exactly.
3. Cost your top 5 sellers with prompt 11, and re-price anything over your target food cost.
4. Run the sales review (prompt 15) and pick one dish to promote and one to rethink.
5. Make the delivery-app version with prompt 18.
6. Check allergens against your recipe sheets, by a person, before anything is printed or uploaded.

Frequently asked questions

Can ChatGPT write my restaurant menu?

Yes, ChatGPT, Gemini or Copilot can write a full first draft of menu descriptions in minutes if you give them each dish's name, real ingredients and your tone. Tell it to use only the ingredients you list and not to invent origins or health claims, then edit the draft so it matches your kitchen exactly.

Can AI work out the allergens in my dishes?

No. AI only sees the ingredients you type, not your stocks, sauces, oils, garnishes or supplier products, so it can miss allergens. Allergen information must come from your recipes and supplier specifications and be checked by a person. Use AI only to format an allergen table you've already written.

How do I calculate menu prices with AI?

Give the AI each ingredient's quantity per portion, your supplier prices and pack sizes, and your target food cost percentage. Ask it to show the working line by line, then the price needed to hit your target. Check the maths, and make clear whether your prices include sales tax or VAT.

What is menu engineering?

Menu engineering is a way of reviewing your menu by sorting each dish by popularity (how many you sell) and profitability (gross profit per portion). You keep and feature dishes that do well on both, re-cost popular low-margin dishes, promote profitable slow sellers, and rethink dishes that do badly on both. AI can do the sorting from your till data.

How long should a menu description be?

Most menu descriptions work best at around 10–15 words: enough to say the main ingredients and cooking method without slowing the customer down. Delivery-app descriptions should be shorter and say the portion size. Keep signature dishes slightly longer if they need explaining.

Is it safe to use AI-generated food photos on my menu?

It's best not to. AI-generated food images often look different from what you actually serve, which can disappoint customers and lead to complaints. Use real photos of your own dishes, even phone photos taken in good daylight.

More help for food businesses: read our AI guide for restaurants and cafés, WhatsApp Business quick replies (including prompts for restaurants), how to reply to negative reviews with AI and AI prompts for Google Business Profile posts. Want 50+ ready-made prompts for menus, reviews, functions and rosters? See the AI Playbook for Restaurants & Cafés.

Get a personalised AI plan for your business free: take the 2-minute quiz.