It really is. And why so often? It is hard to imagine this construct is that prevalent in training data.