Comment by ActivePattern

Because a math proof is generally de-composable into smaller constituent arguments that can be independently evaluated for correctness. So you can audit a sample of these arguments for correctness without reviewing everything. You would then verify that the arguments being true implies the conclusion.