You might try utf8::downgrade on a copy of the string and see if it fails. I don't know whether that's a valid solution, but it sounded good to me.