Free AI for Recipe Scaling and Kitchen Math: What It Gets Wrong

By Austin Little

Ask a free chatbot to turn a recipe for six into a recipe for fifteen and you'll get an answer in two seconds that looks perfectly tidy. Some of it will be right. Some of it will quietly ruin dinner. The difference is knowing which parts of cooking are arithmetic and which parts are physics, chemistry, and food safety wearing an arithmetic costume.

AI disclosure. This page was drafted with AI assistance and edited for Apiary. We don't invent quotes, stats, people, or events.

Every food-safety number on this page comes from an FDA page or the federal FoodSafety.gov site, fetched on October 3, 2026. We tried to fetch the USDA FSIS temperature chart directly and it blocked our fetcher, so we cross-checked the same numbers on the FDA and FoodSafety.gov pages instead. If a chatbot gives you a different safety number, the official chart wins. Every time.

The short answer

Free AI is good at:

• Multiplying and dividing ingredient lists.
• Converting between US measures (teaspoons, tablespoons, cups).
• Converting between Fahrenheit and Celsius.
• Rewriting a recipe into a cleaner format, or a shopping list.
• Explaining why a technique works.

Free AI is unreliable at:

• Knowing how much a cup of your flour weighs.
• Adjusting for a different pan size or depth.
• Adjusting cooking and baking times when quantities change.
• Handling eggs, leavening, and seasoning when you scale.
• Anything about home canning, preserving, or food safety that it didn't get from an official source.

Free AI should never be the final word on:

• Safe internal cooking temperatures.
• How long food can sit out.
• Thawing, cooling, and storing leftovers.
• Allergens, special diets, or medical nutrition needs.

The rest of this guide explains why, with prompts you can copy.

Which free tools work for this

Any general-purpose free chatbot can do kitchen math. For example, Anthropic's pricing page lists a Free plan at $0 with chat on web, desktop, and mobile (usage limits apply).

If you'd rather keep everything on your own computer, Ollama's quickstart describes downloading it for macOS, Windows, or Linux and running a local model without an API key. Check its docs for the current hardware guidance before you download a model. Local models are usually fine for multiplication and unit conversion. They're no more trustworthy than cloud models about food safety.

You don't need a paid plan for anything here. A recipe is short, and kitchen math is light work.

Mistake 1: Multiplying is easy; rounding is where it goes wrong

Scaling a recipe is mostly multiplication. If the original serves 4 and you want 6, multiply everything by 1.5. A chatbot does this quickly. The problems start when the result has to land on a real measuring spoon.

Say the recipe calls for ¾ teaspoon of salt and you scale by 1.5. That's 1⅛ teaspoons. Some chatbots round that to "1 teaspoon." Some round to "1¼ teaspoons." Some leave "1.125 teaspoons," which no measuring set has. None of these is wrong exactly, but if the bot rounds different ingredients in different directions, small errors pile up, and they matter most in baking.

What to do:

• Ask the AI to show the exact scaled amount and a practical measurement, side by side.
• Ask it to express small amounts in a smaller unit. US measures convert cleanly: 3 teaspoons make 1 tablespoon, and 16 tablespoons make 1 cup. So 1⅛ teaspoons is easier to measure as "1 teaspoon plus ⅛ teaspoon."
• For baking, switch to weight if you can (more on that below).

Prompt:

Scale this recipe by 1.5. For each ingredient, give (a) the exact scaled amount, and (b) the closest practical measurement using standard US measuring cups and spoons. Flag any ingredient where rounding changes the amount by more than about 10 percent.

That last line is useful. It makes the bot tell you where it's cutting corners.

Mistake 2: Volume-to-weight conversions that pretend to be exact

Here's a classic. You ask, "How many grams is one cup of flour?" and the chatbot gives you one confident number. But a cup of flour doesn't have one weight. It depends on the type of flour, whether it was sifted, and whether you scooped it straight from the bag or spooned it into the cup and leveled it. Different reputable sources publish different numbers for the same ingredient, because they're measuring different methods.

The same goes for brown sugar (packed or not), cocoa powder, oats, shredded cheese, chopped nuts, and anything that can be pressed down or fluffed up.

Liquids are friendlier. A US cup is 8 US fluid ounces, and water converts to weight fairly predictably. But "fluid ounces" (volume) and "ounces" (weight) are different things, and chatbots mix them up. Ask which one they mean.

What to do:

• If the recipe gives weights, use them and a kitchen scale. Weight is the most reliable way to scale baking.
• If the recipe gives only volume, scale in volume. Don't convert to grams and back just because the bot offered.
• If you do convert, ask the AI what assumption it's using ("spooned and leveled?"), and treat the number as an estimate.
• Look for the original recipe author's own conversions. Many published recipes list both.

Prompt:

Convert this ingredient list from US cups to grams.

Mistake 3: Pan size is about area and depth, not just "use a bigger pan"

When you double a cake recipe, the batter has to go somewhere. Chatbots often say "use a larger pan" or name a pan without checking whether the batter will be the same depth. Depth controls how fast the center cooks.

This is plain geometry, and a chatbot can do it well if you ask it to show the math:

• An 8 × 8 inch square pan has a bottom area of 64 square inches.
• A 9 inch round pan has an area of about 63.6 square inches (π × 4.5 × 4.5). So an 8 inch square and a 9 inch round hold about the same batter at the same depth.
• A 9 × 13 inch pan is 117 square inches, a little less than double an 8 × 8.

Those numbers assume straight sides and the same depth. Real pans have sloped sides and different heights, so treat the math as a starting point.

What chatbots get wrong:

• Recommending a pan with the right volume but a much smaller area, so the batter ends up far deeper and the middle stays raw while the edges burn.
• Forgetting that a deeper pan usually needs a longer, sometimes cooler bake.
• Mixing up the pan's top dimensions with its bottom dimensions.

Prompt:

My recipe uses an 8 × 8 inch square pan. I want to make 1.5 times the recipe. Compare the bottom areas of these pans I own: [list your pans with measurements]. Show the math, and tell me which pan keeps the batter closest to the original depth.

Measure your own pans. Labels and listings aren't always the inside dimensions.

Mistake 4: Cooking time doesn't scale with quantity

This one trips up chatbots constantly. Ask, "If a 3 pound roast takes 90 minutes, how long does a 6 pound roast take?" and a careless answer says "180 minutes." That's not how heat works. Heat has to travel from the surface to the center, and the time depends on the thickness of the food, its shape, the starting temperature, your oven, and more. Doubling the weight doesn't simply double the time.

The same goes for:

• Doubled batters in the same pan, which get deeper and take longer.
• Doubled batches split into two pans, which may take about the original time but can cook unevenly if the oven is crowded.
• Big pots of soup or stew, which take longer to come to a boil.
• Halved recipes in small pans, which can overcook fast.

What to do:

• Treat any time the AI gives as a rough guide to when to start checking, not when to stop.
• For meat, poultry, seafood, eggs, and leftovers, use a food thermometer. The FDA's safe food handling page says that color and texture are unreliable indicators of safety, and that using a food thermometer is the only way to ensure the safety of meat, poultry, seafood, and egg products for all cooking methods.
• For baked goods, use the recipe's own doneness cues (a toothpick test, springy center, golden edges) and start checking early.

Prompt:

I'm doubling this recipe. Don't just double the cooking time. Explain which factors change the cooking time, tell me when to start checking, and tell me what doneness signs to look for. For any meat, poultry, egg, or seafood, remind me to check the official minimum internal temperature with a thermometer instead of giving me a temperature yourself.

That final instruction keeps the chatbot from inventing a safety number.

Mistake 5: Eggs, leavening, and seasoning don't always scale neatly

Eggs

Scale a recipe that calls for 3 eggs by two-thirds and you get "2 eggs." Fine. Scale a 1-egg recipe in half and you get "½ egg." Chatbots will happily write "½ egg" without telling you how. The practical fix: beat the egg, then measure out about half by volume or weight. Ask the AI to give you that method, not just the fraction.

When a chatbot rounds eggs (say "1.5 eggs" becomes "2 eggs") in a baking recipe, the liquid balance shifts. In a forgiving recipe like pancakes, you may not notice. In a delicate cake, you might.

Leavening (baking soda, baking powder, yeast)

Strict multiplication is where many people start, and for small changes it's often fine. But bakers commonly note that leavening doesn't always behave proportionally in very large or very small batches, because pan size, batter depth, and mixing all change. Chatbots rarely mention this.

Salt, spices, and heat

Many cooks scale salt and strong spices a bit below the straight multiple, then taste and adjust. That's a cooking judgment, not a rule. The AI can't taste your soup. Ask it to give you the straight multiple and remind you to adjust to taste.

Prompt:

Scale this baking recipe from 12 servings to 6. For eggs, tell me how to measure a partial egg. For leavening and salt, give me the straight-multiplied amount but flag them as ingredients where I might want to check a recipe written at this size.

Mistake 6: Unit conversion errors that look perfectly confident

These are rarer with modern chatbots but still happen:

• Tablespoons vs. teaspoons. The abbreviations "T," "Tbsp," "tsp," and "t" get confused, especially in recipes you've retyped. Spell them out when you paste.
• US vs. UK vs. metric cups. Cup sizes differ between countries.
• Fluid ounces vs. weight ounces, as above.
• Temperature. The conversion is °C = (°F − 32) × 5 ÷ 9. So 350°F is about 176.7°C, which recipes often round to 175°C or 180°C. For oven baking, a few degrees of rounding is normal. For food-safety temperatures, don't round on your own. Use the official chart, which lists both scales. For example, FoodSafety.gov lists poultry at 165°F (74°C).
• Fan or convection ovens. Some recipes adjust temperatures for fan ovens.

Prompt:

Convert all temperatures in this recipe from Fahrenheit to Celsius. Show the exact conversion and the rounded value most recipes would use. If any temperature is a food-safety temperature (internal temperature of meat, poultry, eggs, seafood, or leftovers), don't convert it. Instead, tell me to look it up on an official chart.

Mistake 7: Food safety (where you shouldn't take AI's word at all)

Here's the line we hold on Apiary: a chatbot is not a food-safety source. It can be right most of the time and still be wrong on the one day it matters. Food safety numbers should come from official pages. Here's what we verified today.

Safe minimum internal temperatures

These come from the FoodSafety.gov chart (last reviewed November 21, 2024, according to the page) and match the FDA's safe food handling page. Use a food thermometer in the thickest part of the food. FoodSafety.gov advises not touching bone, fat, or gristle with the probe.

For precooked ham, the chart says to reheat to 165°F (74°C), with a note that cooked hams packaged in USDA-inspected plants can be reheated to 140°F (60°C). Check the chart for details.

Scaling doesn't change these numbers. A double batch of meatballs has the same safe internal temperature as a single batch. It may just take longer to get there. If a chatbot ever suggests a lower temperature because the pieces are smaller or the batch is bigger, ignore it.

The 2-hour rule and the danger zone

From the FDA's safe food handling page:

• Refrigerate or freeze meat, poultry, eggs, seafood, and other perishables within 2 hours of cooking or purchasing.
• Refrigerate within 1 hour if the temperature outside is above 90°F.
• Keep your refrigerator at 40°F or below and your freezer at 0°F or below, checked with an appliance thermometer.

FoodSafety.gov describes the temperature "danger zone," where germs grow rapidly, as between 40°F (4°C) and 140°F (60°C), and says to keep hot food at 140°F (60°C) or above after cooking if you're not serving it right away.

This matters for scaling: big batches take longer to cool. A doubled pot of chili sitting on the stove "to cool down" can spend a long time in the danger zone. The FDA says to divide large amounts of leftovers into shallow containers for quicker cooling in the refrigerator. When you scale up, plan your containers and fridge space too.

Thawing and marinating

The FDA says to never thaw food at room temperature, such as on the counter. It names three safe ways to defrost: in the refrigerator, in cold water, and in the microwave, and says food thawed in cold water or in the microwave should be cooked immediately. It also says to always marinate food in the refrigerator, and not to reuse marinades used on raw foods unless you bring them to a boil first.

A big batch means more thawing time in the refrigerator. If you're scaling up for a party, start thawing earlier rather than letting a chatbot talk you into a faster "hack."

Reheating

The FDA says to bring sauces, soups, and gravy to a boil when reheating, and lists leftovers at 165°F. When microwaving, it says to cover food, stir, and rotate for even cooking, and to allow standing time before checking the internal temperature.

Home canning and preserving: don't scale with AI

Home canning (jams, pickles, canned vegetables, sauces) is the one area where we'd tell you not to use a chatbot for scaling or recipe changes at all. Safe home canning depends on tested recipes and processing methods, and changing quantities, jar sizes, or ingredients can affect safety.

Allergies and special diets

AI can suggest swaps (oat milk for dairy milk, flax "eggs," and so on), but it can't guarantee a dish is safe for someone with a food allergy. Ingredients and labels vary by brand and change over time. Read the actual labels, and follow advice from the person's clinician.

A safe, repeatable workflow for scaling with AI

Here's the routine we'd use.

1. Paste the recipe cleanly

Type or paste the ingredient list and steps. Spell out units ("tablespoon," not "T"). Include the pan size and the number of servings.

2. Tell the AI the rules

Start your message with something like:

You're helping me scale a recipe. Rules: show exact math before rounding; give practical measurements using standard US cups and spoons; don't guess weights for dry ingredients without saying what assumption you're using; don't give me any food-safety temperatures or holding times yourself — tell me to check an official chart instead; and flag anything that may not scale linearly.

Chatbots follow instructions like these surprisingly well, and they make the output much easier to check.

3. Ask for the scaled recipe and a "check this" list

Now scale it from [X] servings to [Y]. At the end, list every place where I should double-check your work.

4. Spot-check two or three lines yourself

Pick the most important ingredients (flour, liquid, eggs, salt, leavening) and check the multiplication on a calculator. If any is wrong, tell the bot and ask it to redo the whole list.

5. Check your pan

Use the pan-area prompt above, or just measure and do the math.

6. Look up safety numbers on the official chart

For anything with meat, poultry, seafood, or eggs, look up the temperature on FoodSafety.gov or the FDA page. Write it on the recipe.

7. Cook with a thermometer and your senses

Start checking early. Trust the thermometer for safety and your eyes and nose for everything else.

8. Write down what happened

If you scaled a recipe and it worked, note the changes. Next time you won't need the chatbot at all.

Other kitchen-math jobs AI is genuinely good at

Not everything is a trap. Free AI is useful for:

• Shopping lists. "Turn this scaled recipe into a shopping list grouped by store section."
• Batch planning. "I'm making three recipes for a party. Combine the ingredient lists and tell me which ones overlap."
• Cost estimates. "Here are the prices I paid last week. Estimate the cost per serving." You supply the prices. Don't ask the bot to guess them.
• Timing a meal. "These four dishes need to be ready at 6 p.m. Here are their prep and cook times. Work backward and give me a schedule." Check its arithmetic, and remember cook times are estimates.
• Explaining technique. "Why do recipes say to rest dough?" "What does 'fold' mean?" General explanations are a strong suit, as long as they aren't food-safety claims.
• Reformatting. Turning a long blog recipe into a clean card with just ingredients and steps.

When the chatbot and the recipe disagree

Sometimes the AI will "improve" a recipe without being asked. It changes a temperature, swaps an ingredient, or skips a step. If the original recipe came from a source you trust, keep the original. Tell the bot: "Don't change anything except the quantities."

When the chatbot and an official food-safety chart disagree, the chart wins. No exceptions.

When the chatbot and your own experience disagree, the bot doesn't know your oven, your altitude, your pans, or your family's taste.

Prompts to copy

Basic scaling

Scale this recipe from [X] to [Y] servings. Show exact amounts first, then practical US measurements. Don't change anything except quantities.

Rounding audit

Which ingredients in your scaled version were rounded by more than 10 percent? List them with the exact and rounded amounts.

Pan check

My original pan is [size]. I own these pans: [sizes]. For a [factor]x batch, which pan keeps batter depth closest to the original? Show the area math.

Time guidance (safely)

For the scaled version, when should I start checking for doneness, and what signs should I look for? For meat, poultry, eggs, seafood, or leftovers, tell me to check an official minimum internal temperature with a thermometer. Don't give a temperature yourself.

Party cooling plan

I'm making a quadruple batch of [dish]. Help me plan how to cool and store leftovers quickly, using shallow containers, and remind me to check the official 2-hour rule and refrigerator temperature guidance.

Self-check

Review your last answer. Which parts are simple arithmetic, which parts are estimates, and which parts should I verify against an official food-safety source?

The short version

• AI is great at arithmetic and reformatting, shaky at physics and chemistry, and not a food-safety source.
• Make it show its math and flag where it rounded.
• Use weight for baking when the recipe gives it; don't trust a single grams-per-cup number.
• Pans are about area and depth. Measure yours.
• Cooking time doesn't double when quantity doubles. Use doneness cues and a thermometer.
• Safe temperatures don't change with batch size. Get them from FoodSafety.gov or the FDA, not a chatbot.
• Big batches cool slowly. Use shallow containers and respect the 2-hour rule.
• Don't scale home-canning recipes with AI.

FAQ

Can I trust ChatGPT or another free chatbot to double a recipe?

For the multiplication, mostly yes, but check two or three key lines yourself. Don't trust it on pan size, cooking time, or food safety without checking.

How do I halve a recipe that uses one egg?

Beat the egg, then measure out about half by volume or weight. Ask the AI to give you that method rather than just writing "½ egg."

Does a bigger batch need a higher cooking temperature to be safe?

No. Safe minimum internal temperatures depend on the food, not the batch size. Poultry, for example, is listed at 165°F (74°C) on FoodSafety.gov whether you're cooking one breast or ten. A bigger batch may just take longer to get there.

How long can food sit out at a party?

The FDA says to refrigerate perishables within 2 hours, or within 1 hour if it's above 90°F outside. Check the FDA's safe food handling page for the full guidance.

Is there a free AI that doesn't send my recipes to the cloud?

Yes. Local apps such as Ollama run models on your own computer, and Ollama's quickstart says local models don't need an API key. For recipes, privacy usually isn't the big concern. Accuracy is, and that's the same locally or in the cloud.