Start with an independent reference
Before scanning, record what you intend to preserve: all faces, only numbered pages, or a defined section. Count physical sheets separately from retained faces. Covers, foldouts, inserts and intentionally blank backs can make a simple doubled sheet count misleading.
Keep the first unprocessed PDF. A later cleanup stage can change the page count without changing a scanner’s success message. An expected label range can reveal missing first or last labels that a sequence-only check cannot discover.
Read the number that is actually printed
Select a narrow top or bottom band that includes the folio. The tool tries the PDF text layer first, then OCR on the rendered crop when enabled. Dates, chapter numbers and running headers can be mistaken for folios, so inspect the crop instead of trusting a number just because it was extracted.
- Use a full-width band for alternating outside-corner numbers.
- Narrow the region when several numbers compete.
- Review every OCR candidate before relying on it.
- Use output-page=label corrections and retain the ledger of manual changes.
Turn a gap into a focused rescan task
A jump from 12 to 15 is evidence of an unconfirmed interval. If two intervening pages are unread, they may contain 13 and 14. If a book intentionally skips a number, the content may still be correct. Compare the flagged neighboring PDF pages with the physical source, then rescan only the confirmed omission.
A clean sequence is useful evidence, but it is not a content-completeness certificate. Unnumbered inserts and cropped paragraphs require a separate visual review. Save the original file, the page ledger and any corrected scan together.