I wonder if sending jpgs of text rather than posting text in social media messages would gum up the works when they censor communications.
Software CAN scan for characters but when I have scanned for specific meme matches, the software recognizes the picture or that it is a meme but not the specific text.
People can still report “your” posts but if your profile is set to friends only, one of your “friends” has to raise a stink about it.
OCR is getting better but still much less filterable in graphic forms.
It’s also part of the genesis of “lolspeak” - the purposeful misspelling of words to dodge datamining and filters.