If the paper is worth it and does have an original not OCR-ed text layer, it’d better be exported as any other format. We don’t call good things a PDF file, lol. It’s clumsy, heavy, have unadjustable font size and useless empty borders, includes various limits and takes on DRM, and it’s editing is usually done via paid software. This format shall die off.

The only reason academia needs that is strict references to exact page but it’s not that hard to emulate. Upsides to that are overwhelming.

I had my couple of times properly digitalizing PDFs into e-books and text-processing formats, and it’s a pain in the ass, but if I know it’d be read by someone but me, I’m okay with putting a bit more effort into it.