Restaurant calories: how far off the menu is, and how far off you are

By Robert Robinson, founder of Neato Health · September 15, 2026 · 4 min read

The menu says 650. The plate is 900. You guessed 500. Here is what the studies measured, why the errors all run the same direction, and how a photo does better than a guess.

Eating out is where calorie counting breaks. At home you can weigh a portion and read a label. At a restaurant you have a menu number that may be wrong, a plate that may be bigger than the one the number was measured on, and your own eyes, which are worse at this than you think. The research on all three is consistent and a little alarming.

The menu is off by a lot, and not evenly

In 2011, Urban and colleagues bought 269 food items from 42 restaurants, both sit-down and fast food, and measured their actual calorie content in a laboratory by bomb calorimetry. On average the stated calories were close to the measured ones. But the average hid a wide spread: 19 percent of items contained at least 100 calories more than stated, and some contained several hundred more. The errors were larger at sit-down restaurants than at chains, and larger on lower-calorie items, which is the worst place for them: the “light” choices were the ones most likely to be understated.

There is a structural reason. A chain’s menu number is measured once, on a standard portion, in a test kitchen. Every restaurant then makes it by hand with a ladle and a judgment call. The listed number describes the recipe, not your plate.

You are off by more

Block and colleagues asked 3,385 diners at fast-food restaurants in New England to estimate the calories in the meal they had just bought, then compared the estimates with the chains’ own figures. Adults underestimated by 175 calories on average; teenagers by 259. The larger the meal, the larger the miss: for meals over 1,000 calories, a quarter of diners were off by 500 or more. Almost nobody overestimated.

That direction matters. Random error averages out over a month of logging. Error that always runs one way does not; it accumulates into a deficit that exists on paper and not in life.

And it adds up

Nguyen and Powell used national dietary survey data to measure what a restaurant meal does to the day. Eating at a fast-food restaurant added about 190 calories to the day’s intake, and a full-service restaurant added about 190 as well, with more sodium, more saturated fat and more sugar than the same person’s home meals. Two restaurant meals a week at that margin is a pound of fat every two months, from meals that probably felt no bigger than normal.

Why a photo beats a guess

None of this means restaurant meals can’t be logged. It means the logging has to look at the plate rather than the menu or the memory. That is what photo-based food records are for, and their accuracy has been studied for over a decade. Martin and colleagues validated the Remote Food Photography Method, in which people photograph meals and trained analysts estimate portions from the images, against the gold standard of doubly labeled water and found it underestimated intake by about 3 to 6 percent. Written food diaries, by comparison, typically miss by 20 to 40 percent.

Neato’s Snap does the analyst’s job with a vision model: it identifies the foods on the plate, estimates the portion of each from the image, and applies nutrition data for that food. It is not exact, and no method short of a lab is. But it is looking at the actual portion in front of you, which the menu number is not, and it is not subject to the wish that the meal was smaller, which your own estimate is.

How to log a restaurant meal well

• Photograph the plate before you start, from slightly above, with the whole plate in frame. Sides and sauces count; get them in.
• If there is a menu with calories, snap that too. Neato’s menu snap reads the listed number and uses it as a starting point, then adjusts for what the plate shows. The menu figure is a reasonable anchor for chain food and a loose one for everything else.
• Log the drink separately. A large soda, a beer or a glass of wine is 150 to 250 calories that the plate photo does not see and the menu rarely lists.
• Round up when unsure. Given that every study finds the errors run low, a habit of rounding up on restaurant meals corrects the bias the research says you have.
• Edit after. If you left half, tap the entry and cut the servings. The number should describe what you ate, not what arrived.

The bigger point

People who log at home and guess at restaurants have a log that is right five days a week and quietly wrong on the days that matter most. The Block study’s finding that the biggest meals carry the biggest misses is the whole problem in one sentence. A restaurant meal is exactly the meal to photograph, because it is exactly the meal you will misjudge.

Neato’s guides are general wellness information built from the published research cited above. They are not a diagnosis, treatment, or medical advice, and they don’t know your history. For decisions about your health, talk to your doctor.