Typically for me (at least) bad encoded files have been processed to find 'thing' and converted to the Microsoft belief you meant to use the real pair of quote marks they prefer. processed by super-smart software. the worst kind. "I was only trying to help" software.
Cut/Paste of text in from a web browser can trigger this.
I usually see it doing python file read. I have to remember to brute-force open as UTF-8 or else python does what it thinks is needful and marks the stream as binary, and not suitable for use in the str() function suite.