The Celtic-speaking peoples of Britain, Ireland and Gaul left no religious literature of their own, and according to the classical sources that is deliberate: the druids are reported to have refused to commit their teaching to writing, though they used Greek letters for ordinary business.
The consequence is that the earliest descriptions come from conquerors. Caesar is the fullest and he is writing about a people he was in the process of subduing. The classical accounts agree on human sacrifice and on a learned priestly class, and they also copy one another, so their agreement is weaker evidence than it looks.
The insular material is much richer and much later. Irish and Welsh tales were written down by Christian scribes from the medieval period onward, and they preserve figures and motifs that are plainly pre-Christian while presenting them in a framework that is not. Sorting the layers is the central technical problem of the field.
Archaeology supplies what the texts cannot: votive deposits in rivers and bogs, the bog bodies, sanctuary sites, and inscriptions from the Roman period that name local gods paired with Roman ones.
Then there is the invented layer, and it has to be named because it is still in circulation. The Welsh antiquarian Iolo Morganwg produced a substantial body of druidic material in the late eighteenth and early nineteenth centuries that he presented as ancient and had written himself. Much of what the public knows as druidry descends from it.