TBH any document format is going to have the images and graphics and graphics as text inline of the document issue for AI to ingest it. Someone could just as easily stick that graphic, graph or block of text as an image into a word doc or an openoffice doc. For all of these AI would likely need to render the document to an image and then just ingest it all using OCR. Then there's probably even questions about things like graphs and charts. How does AI ingest this, as a raw bitmap to just be redisplayed, or does it analyze it, extract the raw data that the graph or chart represents and store that so it can be graphed or charted out in the future in some different format?