AFAIK, screen-readers read out the full Unicode name for the emoji. Every time. So becomes “Reversed Hand With Middle Finger Extended” to a screen reader. Something to consider the next time you decide to write a clapping hands sentence . (or as screen readers see it, “write Clapping Hands Sign a Clapping Hands Sign clapping Clapping Hands Sign hands Clapping Hands Sign sentence Clapping Hands Sign.”)
And the less said about ASCII-art memes, the better.
The thing is, this isn’t really something I think is on screen-readers to fix. Not on their own, anyway. They’re doing exactly what they’re supposed to do: read the text as written. These “fancy” characters are deep in the Unicode extended character set for mathematical symbols, and were never intended to be used as replacements for standard Latin characters in everyday writing. They’re not just bold or italic letters, they have their own semantic meanings, which is why VoiceOver is treating them this way to begin with (also, for the record, VoiceOver’s reading speed in this person’s tweet has been cranked up pretty high, which I suspect most of its day-to-day users are used to, but it does also strongly emphasize just how unintelligible these symbols can make things).
We have actual semantic tags built into HTML to do things like indicate emphasis in ways that screen readers are totally fine with, but Twitter—and most other social media platforms—don’t provide any means of inserting them into the text of a message. Add a text-formatting bar to the tweet field so you can actually bold and italicize normal ASCII/extended characters, and this problem mostly solves itself.