How the converter works
When we got an air fryer, I searched for a conversion chart and got a different answer from every site. Drop the temperature 25 degrees. Cut the time by a fifth. No, by a third. None of them showed their work, so none of them could say when their rule would fail.
The converter on the front page is my attempt at a better answer, and this page shows the work. I have tried to keep the claims small and the arithmetic honest.
It’s the wind, not the heat
An air fryer is a small convection oven with the fan sitting close to the food. Two things follow from that.
First, still air is a good insulator. Food in a regular oven wears a thin blanket of cooler, moist air, and heat has to cross it to get in. A fan thins that blanket; the harder it blows, the faster heat arrives.
Second, the box is small. The fan is inches from the food, and the air has nowhere to slow down before it comes around again. Together these mean an air fryer set to 350 °F cooks roughly like a plain oven set to 415.
One line of arithmetic
Every conversion on the front page comes from this:
air fryer minutes =oven minutes × 10(oven − fryer − β) ÷ z
Two letters need introducing. β (beta) is the wind, measured in degrees: how much hotter this machine effectively runs than a plain oven at the same setting. My starting estimate for a typical basket model is 63 °F. The letter z sets how strongly temperature trades against time. I use z = 250, which means every 75 °F of extra effective heat cuts the time roughly in half.
A recipe is a dose: “425 °F for 35 minutes” is one quantity of cooking delivered on a schedule, not two separate facts. Change the temperature and the equation rebalances the time so the dose stays the same.
A worked example, the one I test the app against: a recipe at 425 °F for 35 minutes. The app proposes dropping the dial by 25, to 400 °F. The exponent works out to (425 − 400 − 63) ÷ 250, which makes the multiplier about 0.7, so call it 25 minutes. The app itself says 22, with a first check at 16. The difference is a deliberate safety margin, covered in its own section below.
The dial drops by only 25 °F even though the wind is worth 63. That is deliberate: running effectively hotter than the original oven is what produces the crisp. The app checks that the middle still has time to cook through at that pace, and for thick, slow foods it steps the dial down further until it does.
The equation is exact about the trade. What it cannot know is your machine: two air fryers set to 400 °F do not blow the same, so no single formula is accurate for every machine out of the box. That uncertainty lives in one number, β.
Where β comes from
We use 63 °F as the starting β because it is the midpoint of what published recipes solve to. Run the equation backward on oven and air fryer versions of the same dish, holding z at 250, and every savoury category of ordinary thickness (fries, vegetables, proteins from burgers to whole roasts, reheating) lands between 58 and 68 °F. 63 sits within 5 °F of all of them. Food shaped like a board solves far higher; that gets its own setting and its own section below.
A 5 °F miss changes the time by about 4.7%, one minute on a 22 minute cook; a typical home oven thermostat is off by 25 to 50 °F, worth about 40%. That gap is why the app never asks what food you are cooking, only how thick it stands: thickness is the one trait that moves the answer by more than the thermostat noise. The other exception is baked goods, which solve to roughly 32 to 45 °F, so batters and risen bakes are out of scope rather than averaged in.
63 is only the starting point. Your machine shifts it three ways:
Shape matters most: a small basket concentrates the airflow, an oven with racks is gentler, a toaster oven hybrid gentler still. Watts per litre nudge it either way, because more power in a smaller box blows harder. The last line is yours: each rating moves β for your machine. The first “Burnt” on a fresh profile moves it 25 °F, later ratings move it less, and it can never drift more than 60 °F, so one accidental tap cannot wreck a profile. All of this stays in your browser.
One footnote: if the original recipe was written for a fan oven, part of the wind is already spent, so the fan checkbox on the front page gives 25 °F of β back before the math runs.
The floor, tuned by a steak pie
The equation is deliberately about one thing: how fast hot air delivers heat to the surface. Moving heat from the surface to the middle is conduction, and a fan cannot speed conduction up. So the model pairs the equation with a floor: a converted time never falls below a set share of the original oven time, so the middle keeps the time it needs to heat through.
The floor is tuned the same way β is: from real cooking. I first set it at 55% across the board, and fries and wings came out right. Then a Trader Joe’s steak pie, 350 °F for 45 minutes, came out cold in the middle at the floor’s 27 minutes. That pie was evidence about where the floor belongs, exactly as a Burnt or Under rating is evidence about where β belongs.
What the pie taught the floor: hot, fast recipes are limited by the surface, so the wind genuinely shortens them and the floor can sit low. Slow, gentle recipes were always waiting on the middle, so the fan helps them less and the floor must sit higher. The floor now slides with the recipe, from 55% for hot recipes up to 78% for gentle ones, and the same pie converts to 35 minutes today.
Flat, thick, deep
The floor says the fan cannot rush the middle. The mirror statement is just as true: the fan is at full strength when the food is all surface. How much of a cook the wind can speed up is set by how much of the food is surface, and that is a question of thickness. So the one thing the converter asks about the food is how thick it stands, as three pictures you pick by eye.
Thick, the middle picture, is everything described so far, unchanged: β starts at 63 and the floor slides from 55 to 78 percent. It is the right pick for most savoury cooking: fries, wings, chops, chicken pieces, vegetables.
Flat is for food shaped like a board: flatbreads, pizza, bacon, quesadillas. The middle of a flatbread sits a few millimetres from the surface and heats through in a couple of minutes, so nearly the whole cook is surface work, which is exactly the work the wind speeds up. On the flat setting β starts at 108 °F instead of 63. The dial also drops 50 °F instead of 25: browning runs hot and fast on flat food, and a slightly cooler dial widens the window between done and burnt at a cost of only a couple of minutes.
This setting exists because of a field test, the same way the sliding floor exists because of a steak pie. A Trader Joe's flatbread, 450 °F for 12 minutes in the oven, was done in the air fryer in 5 to 6 minutes when the middle setting predicted about 10. Run the equation backward and that cook solves to a β near 135, far above the 58 to 68 band, and published air fryer times for other flat crisp foods solve to the same neighbourhood. The flat setting starts at 108: inside that evidence but held a little short of it, because a time that errs short is the recoverable miss, and your ratings tune it from there.
Deep is for food where the middle is most of the job: lasagna, pot pies, deep casseroles, whole roasts. Here the floor carries the answer at 78 percent of the oven time no matter how hot the recipe runs. Without this setting a hot recipe for a deep dish would be treated as surface work and handed the low floor, and the middle would arrive cold: the steak pie mistake again, at a higher temperature.
Why the time errs short
I do not know your machine’s true β. Before any ratings come in, the uncertainty is about ±30 °F, and three pictures of thickness cannot see exactly how thick your chicken thighs are. When a number is uncertain, the two possible mistakes have different costs: underdone costs three more minutes; burnt costs the whole dish.
So the app quotes a time near the short end of what it believes, roughly the 30th percentile, and pairs it with a check time placed early enough that even a machine running hotter than estimated will not have overshot before you first look. That is why the front page says the time errs a little short on purpose.
What this is for
Every air fryer is different: a different fan, a different box, a different thermostat. No single chart can be right for all of them, and this converter does not claim to be. The short bias touches browning only, never safety; doneness targets live on the food safety page, and a probe thermometer settles any doubt. Everything else in this tool works toward one goal: tuning temperature and time until your machine cooks it just right, and every rating you give gets it one cook closer.