It used to be a problem that you couldn't cut and paste text from TeX-generated PDFs, and I assume this would be a problem for screen readers. I think the dvi->pdf conversion produced a PDF consisting mostly of coordinates and characters (in a non-standard encoding), which PDF readers couldn't put back together into words. This seems mostly fixed on recent pdfs, but I don't know if it's TeX (pdflatex?) or the readers that have gotten smarter.

Funnily, Donald Knuth's AoCP fascicles ship as postscript files, which Preview.app converts to PDFs that can't be copied from.