As long as the underlying DOM uses the correct semantics, there shouldn't really be a great difference. One can write a static page that would be less accessible than a js heavy alternative. Screen readers nowadays work with the rendered page DOM, not the raw html.