Style is the dough; toppings are the identity on top of it. 4,238 of 5,130 recipes name at least one recognizable topping (97 canonical toppings after normalizing the free-text names). Below: what the internet actually puts on pizza, what makes each regional style itself, and which toppings travel together.
The canon
Share of topped recipes that include each topping. Mozzarella, garlic and basil read more like infrastructure than choices.
onion
31%
garlic
25%
basil
25%
oregano
21%
spices
16%
bell pepper
15%
olives
12%
mushroom
11%
pepperoni
11%
sausage
11%
chicken
8%
chili flakes
8%
egg
7%
garlic powder
7%
rosemary
7%
bacon
7%
spinach
6%
parsley
6%
arugula
6%
ham
6%
What makes a style itself
Signature toppings per regional style: the ones a style uses far more than the rest of the dataset does (“3×” = three times the global rate). Shown for the styles with at least 10 parsable doughs and any other regional style with at least 8 topped recipes.
Lift between the top toppings: red = the pair appears together more often than chance (classic combinations), grey = roughly independent, light = they avoid each other. The matrix is symmetric, so only the upper triangle is drawn. Hover a cell for numbers.
Topping profiles as the recipe names them (title or source tag), deduplicated so “four cheese” and “quattro formaggi” count once. A recipe can carry more than one.
Method: 22,850 topping rows were classified. 2,144 name the base (dough, crust, naan, tortilla) and 2,910 are pantry noise (oil, salt, flour, seasoning); both are excluded. Of the rest, 2,103 rows (12%) did not match the vocabulary and are not counted anywhere above: mostly non-English ingredient names and rarities such as ingredients, origano, mango chutney, slivered almonds, molasses, finely grated lemon rind. Lift = how much more often X appears in a group than in the dataset overall; pairs with fewer than 5 shared recipes are left unscored.