Trying to trace the evolutionary history of armadillos is hampered both by the highly fragmentary nature of the earliest fossils, and by some confusion as to how exactly they should be classified. Historically, armadillos were considered to belong to a single family of animals, technically termed the Dasypodidae. For much of the 20th century, this family was placed in a broader group consisting of a number of mammals that either had teeth with an unusually simple structure or none at all. Since this clearly wasn't a primitive feature (in that teeth were already reasonably complex even in the sort of creatures they must have descended from) the reasoning went that they had lost the more usual dental features, perhaps initially to feed on small invertebrates that don't need much chewing.
These mammals, collectively called edentates, could be found in Africa, Eurasia, and both the Americas and, historically, in Europe, too. That seemed a rather broad distribution, but if they really were incredibly ancient, it perhaps wasn't surprising. It was also, as became clear once we had decent molecular and genetic evidence from around the 1990s, wrong.
It turned out that the Old World edentates were entirely unrelated to the American sort, a case of parallel evolution. Instead, we now recognise that the closest living relatives of the armadillo are the anteaters and sloths, and that the three families taken together are only distantly related to... well, anything, really.
Exactly where they do fit is still a matter of debate, although it's clearly somewhere pretty close to the base of the placental family tree. What we can say is that the modern group, now called the xenarthrans, seems to have originated in South America with species only crossing over into the North after the two continents collided towards the end of the Pliocene. Even then, most of them subsequently died out, with only a single species of armadillo currently found wild in the United States.
There's also debate as to whether all living armadillos really do belong to the same family. Since what constitutes a family and what doesn't is essentially arbitrary it's a fairly abstruse argument. A molecular study in 2016 suggested that, as had previously been suspected, glyptodonts - giant armadillo-like animals with a heavy bony shell - were just an unusual form of armadillo, and not merely close relatives.
The upshot of that was that a number of scientific papers written since 2016 have split the traditional armadillo family into two. The Dasypodidae itself still includes the animal that's so familiar to inhabitants of the southern US, along with its closest relatives. A second family, the Chlamyphoridae, includes the glyptodonts along with a bunch of living South American species, some of which are pretty odd-looking even by the standards of armadillos. Even so, not all modern scientists necessarily use this scheme... and since the two families are each other's closest relatives, there's nothing to say that they have to.
Nonetheless, whether they constitute one family or two, armadillos (in a sense that includes glyptodonts and the rather similar-looking pantatheres) are a single evolutionary group, which makes it reasonable to ask what the first ones looked like.
The oldest known armadillo fossil belongs to an animal called Riostegotherium. This lived in southern Brazil a little over 50 million years ago... and that's almost the only thing we know about it, given that all we have are a few individual chunks of bony scale. Unfortunately, much the same is true of most other early armadillo fossils, which tend to be fragmentary at best.
Most, but not all. Just two species of Eocene armadillo are known from reasonably complete remains. The older of the two is Lumbreratherium, which was first described in 2017 on the basis of a fossil found in northwest Argentina. This was joined last year by Pucatherium, previously known only from a few pieces of carapace. Those pieces had been distinctive enough to indicate that it was likely a close relative of the slightly older species, and, while the complete fossil dates to around 40 million years ago, some of the isolated pieces found elsewhere suggest that it may have lived on until as late as 35 million years ago, almost at the end of the Eocene.
The Pucatherium fossil shows an animal roughly the size of the nine-banded armadillo (Dasypus novemcinctus), the living species found everywhere from the southern US to Uruguay, but it clearly belongs to an entirely different lineage. If we do divide living armadillos into two families, this and Lumbreratherium belong to neither, representing an early branch that died out early on.
We already know, from some more fragmentary remains, that the two living families (or subfamilies, if you prefer) already existed at this point, so these Eocene species can't be the direct ancestors of modern armadillos, but they do retain some decidedly "primitive" features. For example, the carapace of Pucatherium consists of no less than 36 bands of scales, with no sign of the solid shields that the nine-banded armadillo has over its shoulders and hips. Since the scales (more technically "scutes" or "osteoderms") were made of solid bone, they would have provided armour to the animal, but it would have been much more flexible than in most living species, something that presumably had both advantages and disadvantages.
The rest of the skeleton shows a mixture of the features we'd expect to see in the two living groups, and that the animal was more heavily built than a nine-banded armadillo, with relatively strong and thick bones. The teeth were also different from those of modern species. As noted above, armadillo teeth are relatively simple in structure, all appearing essentially identical along the length of the jaw, rather than the usual division into incisors, molars, and so on seen in most mammal species. But, while Lumbreratherium had no incisors, it did have teeth that look like canines, separated by a gap from the more typical armadillo-type teeth behind them.
The history of armadillos must stretch at least ten million years further back in time than these Argentinian animals, and almost certainly much further than that. And it's unlikely that the real ancestors of the group looked exactly like these, more modern, specimens. But they do give us a partial glimpse into what at least some of these strange animals were like in the early days of their evolution.
[Photo by "Mwcolgan8", from Wikimedia Commons.]