You have the same brownie open in three tabs and three different calorie counts in front of you. Most of the time none of those sites made an error. They made different decisions, each one defensible on its own, and the decisions stack.
Five of them account for almost all of the gap, and four are visible on the page if you know where to look. Our own workings, including where they are weakest, are written up in how recipe macros are calculated, and the short version of what follows is that the method tells you far more than the number does.
One: how many pieces the pan got cut into
Calories per serving is a division, and the divisor costs nothing to change.
Our black bean brownies come out of an 8x8 pan cut sixteen ways, so a serving is a two-inch square at 110 calories. Cut that identical pan twenty-four ways and the same batter prints in the seventies. No ingredient changed and no one lied. The serving got smaller.
This is the biggest single source of disagreement between two sites and the easiest to check, which is why it has its own article. Read the yield line before the calorie line, every time.
Two: whether anyone weighed anything
A cup is a volume and flour has no fixed density. Spooned in gently and levelled it lands near 120 grams. Scooped straight from the bag it can come out at 140 or 150, because scooping packs it.
That is a swing of about a quarter on the heaviest dry ingredient in most bakes, and it happens before a calculator sees the recipe. Two sites publishing the identical ingredient list can be a hundred calories apart across a pan purely on how the person entering it assumed a cup was filled.
Nut butter does a smaller version of the same thing. Nobody levels peanut butter, and a heaped tablespoon against a level one is a real difference.
Three: how the calculator treats the sweetener
For a sugar-free dessert, this one is capable of producing a gap bigger than all the others combined, and it is invisible on the page.
Sugar is counted at 4 calories a gram. Allulose is counted at about 0.4 and erythritol at about 0.2, because neither is metabolised for energy the way sugar is. Plenty of food databases have no separate entry for either, so an ingredient line reading "1 cup allulose" gets matched to plain sugar and counted at the full rate.
Our sugar-free marshmallows carry 10 grams of carbohydrate each, nearly all of it allulose, and they print 10 calories. Run that same recipe through a calculator that treats allulose as sugar and the arithmetic lands somewhere above 40 a marshmallow, four times the figure, with not one ingredient changed. Neither number is fraud. One convention says a sweetener your body does not use for energy still counts as energy, and the other says it does not. Why there is no zero-calorie dessert at all is the longer version of what that marshmallow is doing.
So when a sugar-free recipe shows up with a number that seems far too high for what is in it, look at the sweetener first. The mechanism behind the low figure is explained properly in why allulose is not a sugar alcohol. The carbohydrate line splits on the same convention, which is why two net carb figures can disagree and both be honest.
Four: which entry in the database got matched
Every automatic nutrition tool works by matching text to database entries, and generic entries are averages of products that differ from each other.
Peanut butter runs roughly 90 to 100 calories a tablespoon depending on the jar. Protein powders can sit 30 calories apart per scoop, with scoops that are not the same size. Sugar-free chocolate chips are a category rather than a product. Any one of those is noise. Four of them stacked in one recipe is a number that has drifted somewhere real.
Five: what stayed outside the count
A calculator reads the ingredient list. It does not read step six and it cannot see the photograph.
The drizzle described in the instructions, the oil the pan was greased with, the berries in the styled shot, the yogurt in the line that says serve with. If it is not in the ingredient list it is not in the count, and the two lists are not always the same list. Optional ingredients are the awkward case, because they live outside the arithmetic and inside the picture that sold you the recipe.
How to tell which one to trust
Four things separate a number you can lean on from one you cannot, and all four are visible in about twenty seconds.
Does it state the yield and the pan? Sixteen brownies from an 8x8 is a claim you can picture. "One serving" is not.
Does it print fat and sugar as well as calories? Leaving either one off is the oldest way to make a dessert look better than it is, and a site that prints all of it has given up that option.
Does it give any weights? A recipe with grams beside the cups has removed the largest error in the whole process.
Does the arithmetic close? This is the check we run on our own corpus and it takes a minute. Protein and carbohydrate run about 4 calories a gram, fat about 9. Multiply out the macros the recipe printed and add them up. The calorie figure should land at or a little under that sum. Above it is impossible and means something is wrong.
Our black bean brownies print 5 grams of protein, 14 grams of carbohydrate and 5 grams of fat, which multiplies out to a ceiling of 121, against a printed 110. Comfortably inside, and the small gap is roughly what a sweetener sitting in the carbohydrate line but not the calorie line would predict.
That check only catches a number that is arithmetically impossible. It cannot catch one that is wrong and still adds up, and we have found one of those on our own site, which is in the method post. Passing it is a floor rather than a certificate.
The ten-minute version, if the number really matters
Take the two or three heaviest ingredients and recompute only those. In most desserts the flour, the fat and the sweetener are the great majority of the calories, and everything else is rounding.
Weigh your flour rather than scooping it. Use the actual jar of peanut butter in your cupboard rather than a generic entry. Divide by the number of pieces you genuinely cut, not the number the recipe hoped for, because sixteen brownies leaving the pan as fourteen makes every one of them bigger than the label claims.
You will end up with a number that is right for your kitchen, which is the only version that was ever going to be right for you.
Ours are estimates too
Every figure on this site is an estimate and I would rather say that plainly than let the six printed numbers imply a precision they do not have. Most of our recipes still give volumes rather than weights, which is the weakest link in our own process and the thing we are working through.
What we do control is the honest part: a serving is a piece a person would actually eat, and we print fat and sugar next to calories so the number cannot flatter itself. The reasoning behind the portions is in why healthy dessert servings feel small, and if you want to see the standard applied across a whole set, the nine in our sugar-free desserts under 60 calories are all sized that way. The free sampler on this page carries a few of them with the same numbers printed.
If a calorie figure matters to you for a medical reason, the number worth trusting is the one you calculate from the brands in your own kitchen, measured against the targets your care team gave you. Nothing here is medical advice.
The short version
Check the yield, then the sweetener, then whether anything was weighed. Three sites can be honest and still disagree by a third, and the one to trust is the one that showed you enough of its working for you to disagree with it.
