PDFs don’t always contain actual text. Sometimes they just contain instructions to draw the letters.
For that reason, IMO rendering a PDF page as an image is a very reasonable way to extract information out of it.
For the other formats you mentioned, I agree that it is probably better to parse the document instead.