Currently, only PDFs with raw text in them get properly OCRed and added to the search index. We should support PDFs that are just images or contain images.