It sets the interior. A roux sauce firms as it cools inside a hot dish instead of a bowl; a custard sauce is not a sauce until the oven's ambient heat sets the egg, the same way a quiche sets. Neither happens on a stovetop.
Not two recipes. One sauce route, with a lid on it.
Baked mac and cheese
Our largest finish, held for 52 of 117 recipes. Most of what makes a baked mac and cheese what it is was already decided before it went in the oven.
Every recipe site treats “baked” as a style. It's a step. The sauce underneath it is built one of six ways, and 43 of these 52 recipes use just two of them, roux or egg custard. The oven changes three things on top of whichever route a recipe picked: it sets the interior, it dehydrates a top layer into something brown, and it keeps thickening a sauce that already looked right in the pan.
52Baked recipes
28Use a roux
15Use egg custard
36Add a topping
The crossover nobody else holds
Which sauce route the 52 baked recipes actually use
“Baked” is an oven step applied on top of a sauce decision, and the decision is not evenly spread. See the full breakdown of all six routes on roux or no roux; this is just the slice of it that goes in the oven.
| Sauce route | Recipes | Share of baked |
|---|---|---|
| Roux | 28 | 54% |
| Egg custard | 15 | 29% |
| Evaporated milk | 4 | 8% |
| No-cook | 4 | 8% |
| Sodium citrate | 1 | 2% |
| Starch slurry | 0 | 0% |
Roux and egg custard together account for 43 of these 52 recipes. Starch slurry, a route that shows up elsewhere in the corpus, does not appear in the baked cohort at all. Pick a baked recipe and you have picked, almost always, one of two sauces: a stovetop roux carried into the oven, or a Southern-style egg custard set by it.
What the oven actually does
Three things baking changes, on top of the route
Bake a mac-and-cheese sauce looser (higher liquid) than you'd want it at the table, because pasta keeps absorbing liquid in the oven -- a sauce correctly saucy at assembly reads tighter and drier after 25-30 minutes uncovered.
Cited: cypasta.com, specific to the baked format
Undercook the pasta by 1-2 minutes short of stovetop al dente before baking, because continued gelatinization and water uptake in the oven finishes the cooking -- skipping this produces mush by the time the bake is done.
Cited: cypasta.com
Our own tags back the second claim up directly: 13 of 52 baked recipes are tagged as undercooking the pasta before it goes in, a step that shows up far less in recipes that never see an oven. A sauce that reads correctly thin in the pan is not a mistake; it is what a baked sauce is supposed to look like fifteen minutes before it comes out grainy or gluey instead of gone dry. See why sauce turns grainy for the failure mode on the other side of that line.
The other thing an oven does that a
stovetop can't
The browned top, and how the 52 baked recipes build one
Dry heat above the dish dehydrates whatever sits on top of it into something brown and crisp. A pot on a burner has no equivalent surface exposed to air.
| Topping | Recipes | Share of baked |
|---|---|---|
| Breadcrumbs | 15 | 29% |
| More cheese on top | 15 | 29% |
| Crumbs and cheese | 6 | 12% |
| Bare | 16 | 31% |
Top with buttered panko breadcrumbs and bake or broil until browned for a crunchy top.
10 cooks reported this
Broil the dish for a few minutes at the end, instead of relying on a long bake, to brown and crisp the top.
6 cooks reported this
What actually cleared as real
One difference the corpus can stand behind
Not a ranking of baked against every other finish: a two-sample test run against each one, keeping only what survived. Baked names one, on saltiness, not on stability or creaminess.
Stovetop reads saltier than baked.
Measured across 16 reviews of stovetop recipes and 21 of baked ones
Every other pairing, baked against stovetop on stability, baked against slow cooker on anything, is one this corpus does not have enough separation to call.
So, tonight
If you're baking it, decide the route first
The default
28 recipes
Roux, carried into the oven
The most common route in the baked cohort. Build the sauce looser than you want it at the table; the oven tightens it.
What to do: undercook the pasta by a minute or two short of stovetop done, since it keeps cooking under foil or in an open dish.
The Southern route
15 recipes
Egg custard, if you want a real set
A custard-based baked mac is not fully a dish until the oven sets the egg. Pulling it early leaves it loose in a way a roux sauce would not be.
What to do: give it the full bake time even if the center still jiggles slightly; it firms in the last few minutes and again as it rests.
Wants a stovetop finish instead
Skip the oven for a route built to be eaten wet
No-cook and evaporated milk sauces show up in the baked cohort too, but they are a smaller share of it than of the corpus overall. See the stovetop cohort for what these routes look like without a lid.
What to do: if speed is the point, stop after the sauce comes together and skip the bake entirely.
Honesty
What this page cannot tell you
A cohort, not a controlled comparison
These 52 recipes are different recipes by different authors, not one sauce baked and left unbaked to compare. Reading the route table as cause and effect would claim more than the data supports.
Only one measured difference names baked
Of every pairing the finish facet could test, exactly one cleared with baked in it, on saltiness. The claims about setting and thickening come from cited literature and our own process tags, not from a fitted score.
Most recipes don't flag pasta doneness
39 of 52 baked recipes never say whether the pasta was undercooked going in. The undercooking pattern is real in the recipes that do say, but it is not the majority of the cohort.
No fitted lever for salt or melt
Salt percentage and melt index carry no fitted vector in this corpus's model. We can show what the baked cohort does; we cannot show that doing it caused an outcome.