We are now in a strange situation where selecting text on images works better than selecting text in PDFs