PDF files have similar issues. Acrobat can attempt to read them out loud, but unless the PDF author adds extensive annotations, Acrobat has no idea about the "structure" of the document.
For example, headers are just more text. Tables and figures that "float" to the top of a page, or in any other way interrupt the text, get read in the visual, not the logical, order. Images get "read" as a stream of binary data. Equations get mangled horribly. Etc. etc.
LaTeX, much like HTML, allows you to use either stylistic markup ("put this bit in a bigger font") or semantic markup ("mark this bit as a headline").
The PdfLaTeX typesetting engine can embed some semantic information into the PDF. However, it doesn't do everything. For example, I believe it does not handle the issue with "floats" that I mentioned [1].
I don't know whether there are TeX-aware tools for "reading" a LaTeX input file out loud (which is what I think you're asking about). But if there was such a tool, it might well be easier to understand, than listening to a dumb reading of the resulting typeset PDF.
It is possible to go into an existing PDF file and mark up a lot of these things, specifically in order to help the reader along. The webmaster where I work often has to mark up PDF files (usually generated from MS Word documents, rather than from LaTeX). I've never done it, though, so I don't know what's involved. My impression is that it's pretty painful.
[1] To make the issue with floats more explicit. You can tell LaTeX to add a figure, and to "float" it to the top of a page. The typesetting engine will do so. However, suppose the resulting PDF file ends page 4 with "The figure clearly shows..." and then starts page 5 with the figure, followed by "... two trends in the data." Then when the document gets read out loud, the reader will say "The figure clearly shows", then it will go into a spastic literal reading of the figure's contents, then it will finish up with "two trends in the data."
For example, headers are just more text. Tables and figures that "float" to the top of a page, or in any other way interrupt the text, get read in the visual, not the logical, order. Images get "read" as a stream of binary data. Equations get mangled horribly. Etc. etc.