Thanks for looping me in. I'm not subscribed to emacs-orgmode --- so feel free to forward if you find the thoughts below materially useful.
As a long-term org-mode user --- and an even longer term TeX user: here are some thoughts: 1. Accessibility as word used in isolation has now become mostly meaningless, to be concrete one has to ask "Accessibility to whom"? 2. So in the following, everything I say is with respect to users with visual impairments. 3. It's incorrect to define "Accessibility" in terms of a specific user access tool or technology -- that usage is marketing jargon for a specific Access Solution like a screenreader --- so I refrain in general from defining this in terms of Screenreaders. With those meta-thoughts out of the way: A: Org-generated documents are mostly well-structured documents, and not in general a user-interface e.g. (WebApp); so with regard to export -- either HTML or LaTeX/PDF --- ARIA is mostly irrelevant. B: The LaTeX->PDF pipeline *can* produce tagged PDF with respect to document structure eg. Sectioning, but only if you use pdflatex or pdftex i.e. LaTeX/Tex->dvi->[ps]->pdf is lossy with respect to structure present in the markup; this is a short-coming of DVI which predates the thought of document structure making it through to the output. C: pdftex and pdflatex were built in the late 90's by a student in Prague (Hanu Than? from memory) --- only reason I know this is that I got Adobe to fund that project when at Adobe in the 90's. It's a very good piece of work that essentially uses PDF directly as the "Device Independent" format rather than the original dvi. DVI as designed in the 70's was device-independent for the time, ie it did not hardwire printer controls and could be mapped to various print mechanisms. For the 90s, by which time Document Structure meant a lot more than being some version of inkjet printer driver independent, the afore mentioned project used PDF as the Device-Independent format --- and leveraged the Tagged PDF bits from PDF 1.4 to achieve the result. D: All that said, it is likely still easier to go from org->HTML directly and produce content that is easier to machine-process --- rather than go through one more level of indirection via LaTeX and PDF; however there may well be additional constraints in a publication workflow, e.g. publisher wants to only publish final-form -- and yes, in this case, HTML and PDF are both final-form. E: Finally, note that in (D) I said "machine processable" not "Accessible"; machine-processable is a pre-requisite to "repurpose " what you publish, and making that result usable by different user communities is a direct consequence of suche machine-processability. Hope this helps. -- --Raman Thanks, --Raman(I Search, I Find, I Misplace, I Research) ♉ Id: kg:/m/0285kf1 🦮 -- Thanks, --Raman(I Search, I Find, I Misplace, I Research) ♉ Id: kg:/m/0285kf1 🦮