Comment by jncfhnb

Text that is _intentionally constructed_ to fool computers but not humans is obviously out of scope. But they’re generally easily solved with OCR these days anyway.