It keeps the PDF and adds an OCR layer:
There are other options:
But I prefer PDF.
I don’t disagree that being more specific with the criteria may be better, I’m just pointing out that it works as well without it. Kind = PDF resulted in twice as many duplicates (4) the one time I tested it so far. I did find a firmware update and have been able to direct scan on the first attempt after updating, But the multi-page issue is still there for some reason and that’s a deal breaker. Not sure if the reliability is improved yet.
Testing again:
1 copy in Water Bills group, 2 copies in the trash.