[OfSanguineFire]

Wouldn't this require authors to use semantic tags instead of visual-presentation ones? That is, a tag that specifies that what follows is code instead of just \texttt{}, a tag specifying that what follows is e.g. the title of a book instead of just \textit{}, etc. The TeX engine itself, as it processes the source into a PDF, cannot know what the original author meant.

[uxp8u61q]

This isn't about text formatting. It's about things like reading order, alternative text for figures, or even just making clear to the PDF engine that two consecutive words are part of the same paragraph (which isn't even the case by default!). Take a look at this: https://www.overleaf.com/learn/latex/An_introduction_to_tagg... As GP said, it's really shameful that, in academia, we just ignore this kind of stuff. I tried making my articles' PDFs accessible and failed miserably.