A 90-calorie dessert claim is usually a true number attached to something that is not really a dessert, or a true number that quietly left several ingredients out of the count.
The version everyone knows about is the serving size, and we have already written that one at length, so this post is about the other three, which are less discussed and easier to hide. The numerator, the database underneath the calculator, and the fact that a headline number is chosen before it is calculated.
None of this requires anyone to lie. That is the part worth sitting with. Almost every wrong calorie count on the internet was produced honestly by somebody who typed real ingredients into a real tool.
One: the divisor, briefly
Calories per serving is a division problem, and the divisor is the only term anybody can change for free.
Our black bean brownies are 110 calories a square, cut sixteen to an 8x8 pan, which is a two-inch square and a brownie you would recognise as one. Cut the identical pan into twenty-four and the per-piece figure falls by a third without a single ingredient changing.
That is the whole mechanism and the linked post covers it properly, including how to spot it in about ten seconds. The rest of this is about the top of the fraction.
Two: what never made it into the count
A nutrition calculator reads the ingredient list. It does not read the instructions, and it does not look at the photograph.
So anything living outside the ingredient list is free. The chocolate drizzle described in step six. The scoop of ice cream in the styled shot. The "serve with fresh berries and a spoon of yogurt" line under the recipe. The tablespoon of oil you were told to grease the pan with. Chips listed as optional, which are optional in the same way that the frosting on a cupcake is optional.
Optional ingredients have a habit of living outside the nutrition line while living inside the photograph, and the photograph is what sold you the recipe.
There is a version of this that is close to invisible. A recipe that reads "sweetener, to taste" has removed a whole ingredient from the arithmetic and handed the decision to you. If the sweetener is allulose or monk fruit that is genuinely near zero and it does not matter. If it is honey, it matters a great deal, and the recipe has not told you which assumption it made.
Three: the database underneath the calculator
Every automatic nutrition tool works by matching your ingredient strings to entries in a food database. Both halves of that can go wrong.
"1 tablespoon peanut butter" resolves to a generic entry, and generic entries are averages of products that differ. "Sugar-free chocolate chips" is not one product; it is a category with a wide spread. Protein powders vary by twenty or thirty calories a scoop depending on the brand and how much of the tub is filler. None of these is a scandal on its own, and four of them stacked in one recipe is a meaningful drift.
The bigger one is volume against weight. A cup of flour is anywhere from about 120 to 150 grams depending on whether you scooped it or spooned it, which is a swing of roughly a quarter on the heaviest line in most bakes. Almond flour is worse, because it compacts. A recipe that gives volumes only has built in an error bar it never mentions, and the error bar is larger than the difference between most competing "low calorie" claims.
We wrote up how we calculate the macros here for the same reason, which is that the method matters more than the number and almost nobody publishes theirs.
Four: the number was chosen first
This is the one that tells you the most, and you can spot it from the outside.
Look at a site's numbers in bulk. If a dozen unrelated recipes all land on 90, 95 and 100, those figures were not returned by a calculator. Real arithmetic produces 87 and 113 and 142. Round numbers, repeated across a whole site, are targets rather than results, and the way you hit a target is by adjusting the servings count until the arithmetic cooperates.
The order the recipe was built in is the thing being revealed. Pick the serving first and calculate the number afterwards and you get whatever you get. Pick the number first and solve for the servings and you get a pan of brownies cut into twenty-four dominoes.
A check that takes ten seconds
You do not need the recipe's tool. You need the two or three heaviest ingredients and a rough floor.
Fat is where the calories are. Oil and butter run about 120 calories a tablespoon, nut butter roughly 190 for two tablespoons. Almond flour is somewhere near 160 a quarter cup, chocolate chips around 70 a tablespoon. Those are approximations and approximations are all this needs.
So: a recipe claims twelve servings at 90 calories each. That is 1,080 calories for the entire batch. If the ingredient list contains a cup of peanut butter, that single ingredient is somewhere near 1,500 calories on its own, before the flour and before anything else in the bowl. The claim and the ingredient list cannot both be right, and it takes ten seconds to know it.
The check does not tell you which number is wrong. It tells you that one of them is, which is usually enough to stop you making it.
Where this leaves us
Fair is fair, so here is where our own numbers are least flattering.
Our Crumbl-style cookie is 315 calories with 20 grams of fat, and it is one large frosted cookie. A site playing the game would call it two servings and publish 158, or leave the frosting outside the ingredient list, and either move would produce a much better headline for the same object.
Our protein peanut butter cups run 295 calories a serving with 23 grams of fat. Peanut butter and chocolate is a fatty combination and no amount of sugar-free labelling changes that. Fat is where a sugar-free dessert usually hides, which is the same reason we print the fat line rather than leading with calories alone. If you want the full version of that argument, it is in how to read a sugar-free label.
And a nutrition line with exactly one number in it is not being brief by accident. Somebody decided which figure flattered the recipe.
What to do with a claim
Read the servings line before the calorie line, and picture the actual piece.
Then find the fat and protein figures, and if they are missing, treat the calorie number as marketing until proven otherwise. Then, if the headline still looks impossible for what is in the bowl, do the floor check on the two heaviest ingredients.
Most recipes are fine and most people writing them are not trying to fool anybody. The point is narrower than that: a calorie count is only as good as the serving it is attached to and the ingredient list it was built from, and both of those are sitting right there on the page for anybody who reads them in the right order.
