The earliest known depiction is in Bellifortis, a manuscript on military engineering compiled by Konrad Kyeser around 1405. It shows a locking girdle, described as worn by the women of Florence, among a collection of devices that includes items nobody has ever thought were real, in a work whose author was plainly enjoying himself. Reading a joke in an engineering treatise as documentary evidence of a practice is the first move in the whole history of the belief.
From there the motif travels through satire. Early modern writers use the belt as a comic figure for jealousy, in the same register as the cuckold horns, and the joke depends on the device being absurd rather than familiar. Only in the eighteenth and nineteenth centuries does the satire harden into a supposed fact, in a literature that wanted a Middle Ages of cruelty and superstition against which the present could be measured. The crusading husband was added at that stage, and no medieval text supports it.
What the object documents, then, is a nineteenth century idea of the twelfth, and the process by which such an idea acquires evidence. Albrecht Classen traced the sequence in detail, and the conclusion holds up: the belt is a myth with an unusually complete paper trail, which makes it more useful to a historian than a real device would have been.