Documents are sometime OCR and sometime not, depending on the source we fetch them from.