Screen readers that are limited to structured inputs can't be the end-all final solution right? It would be shortsighted to set this tech in stone with legal precedents. Do we seriously want to require every blind person to be fluent in reading html/React as the default way to consume sites?
We should be trying to improve the tech so they can consume the world just like sighted people. There are already apps that can use computer vision to caption and describe images. It can even connect you with real people if the AI sucks.
I think pouring investments into those technologies might be cheaper relative to the amount of work added to the industry as a whole.