How to Track Calories at Restaurants (Without Ruining Dinner)

Restaurant meals are where calorie tracking usually falls apart. There is no label, the database entry is somebody's guess from 2017, and you are not going to weigh your risotto in front of other people. Here is a method that works in the real world, along with the reason it does not need to be perfect.

The short version

• Log it before you eat or immediately after. Memory degrades fast and rounds downward.
• Estimate the cooking fat. It is the single largest thing people miss, often 100 to 300 calories per plate.
• Chains publish real numbers. Use them, they are legally required to be accurate.
• For independent restaurants, describe the dish to an AI tracker rather than hunting for a database match.
• Round up, not down. A deliberate over-estimate beats an optimistic one.
• Being 15% out on two meals a week barely matters. Not logging them at all does.

Why Restaurant Food Is Harder Than It Looks

Three things make restaurant meals difficult to log, and only one of them is portion size.

Fat you cannot see. Restaurant kitchens cook with far more butter and oil than home cooks do, because it tastes better. A vegetable side that reads as 80 calories at home can be 250 in a restaurant. This is the largest and most consistently underestimated variable.

Portions that do not match any standard. Database entries assume a serving size someone decided on. Your plate did not consult it.

Sauces, dressings and finishes. A dressing on a salad can double it. A pan sauce is mostly butter. These get logged as zero or forgotten entirely.

The Method

1. If it is a chain, use their numbers

Large chains publish nutrition data and, in many markets, are legally required to. This data is far better than any crowd-sourced database entry. It is worth checking before you order rather than after, because the differences between menu items are often enormous. We keep a set of pages for the big ones, including McDonald's, Burger King, Chipotle, and Starbucks.

2. If it is independent, describe it in words

Do not scroll a database looking for "chicken tagine". You will find six entries with wildly different numbers and pick one at random, which is worse than estimating. Describe the dish instead: what protein, roughly what size, what it was cooked in, what came with it. A tracker that accepts plain language will produce a better estimate from that description than a mismatched database row ever will.

3. Estimate portions with your hand

You always have this tool with you and it scales roughly with your body size.

4. Add the invisible fat

Unless the dish was steamed, grilled dry, or served raw, add one to two tablespoons of oil or butter to whatever you estimated. That is 120 to 240 calories, and it is the correction that turns a systematically low estimate into a roughly accurate one.

5. Log it at the table

The photo takes two seconds and nobody notices. Logging from memory the next morning is where accuracy dies, because recall shrinks portions and deletes side dishes.

Ordering Choices That Make Tracking Easier

You cannot control a restaurant kitchen, but you can pick dishes with fewer unknowns.

• Grilled or roasted proteins have a much narrower range than anything fried or in a sauce.
• Dressing and sauce on the side converts an unknown into something you control and can see.
• Simple constructions beat complex ones. Steak, potatoes and greens is easy to estimate. A layered curry is not.
• Shared plates are a trap. If you are splitting, decide roughly what fraction you ate before you start, not after.

Why Close Enough Is Genuinely Enough

This is the part people get wrong, usually by giving up entirely. Suppose you eat out twice a week and your estimates are 20% too low on both meals. Across a week of roughly 17,000 calories, that error might be 400 calories, or about 2%. Well inside the noise of daily weight fluctuation, and easily corrected by watching your trend over a month.

Now suppose you skip logging those two meals because you cannot do it precisely. That is potentially 1,800 unlogged calories, or 10% of the week, and it is invisible. You end up wondering why the maths is not working.

The whole game is consistency, not precision. An imperfect log every day beats a perfect log four days a week, and it is not close.

The bottom line

Track restaurant meals by using published data for chains, describing the dish in plain words for everywhere else, estimating portions with your hand, adding a tablespoon or two for the fat you cannot see, and logging at the table before you forget. Then let it be approximate. Bento Bunny is designed for exactly this situation: photograph the plate or describe it in a sentence, and the AI estimates calories and macros in seconds without you searching a database or excusing yourself to weigh anything. It is free to download, and on iOS 26 and later it runs on-device so the photo of your dinner never leaves your phone. If fast food is your regular challenge, see fast-food meals under 700 calories.