We need a browser extension that treats the rendered page as an image, then runs OCR over it, then converts that to something where text can be selected.
Pros: 1. safer (what you see is what you select), 2. also works with images, 3. all text can be selected
This is roughly what reader mode is, no? Safari ships it out of the box, although it's very hit or miss as far as my experience with it goes. But I like the idea.
Pros: 1. safer (what you see is what you select), 2. also works with images, 3. all text can be selected