How PaperTotal reads a document

1. The file is checked, and your original is stored as it came.
2. An image is sized for reading and a thumbnail is made.
3. The text is read. A PDF with a text layer needs no OCR.
4. The merchant, date, amounts, tax and line items are picked out and checked against each other.
5. A category is chosen and possible duplicates are looked for.
6. The expense files itself, or waits under Review if something needs a second look.

Under Settings, General, Reading model decides what reads documents in the workspace from then on. Standard uses OCR and rules, and Claude reads a second time when they are unsure. Rules only sends no page or text to Claude, and anything unsure waits for a person. Always a second reading has Claude read every receipt, invoice and bill as well, which takes longer.