> They could put the e.g. inside a dfn tag or teach the screen readers how to read it.

I'm afraid not. AFAIK, screen readers don't apply any special processing for DFN tags. Speaking as someone who's using a screen reader to read this thread and proofread this comment, I'm not sure I'd want them to either, and I hadn't even heard of them until today.

Also, screen readers are able to speak through several, possibly hundreds, of different speech engines, so you would have to teach each of those engines to deal with these corner cases separately. Given that many of them were developed ten plus years ago and are now no longer under active development despite continued widespread usage, that's not really a realistic goal.

N.B. AFAIK is spoken by my screen reader/speech engine combo as "uh-fake". As a blind software developer, I have much bigger problems to worry about than memorising how my software speaks abbreviations. I of course don't speak for everyone though and clearer communication is always an excellent goal.