PDF to CSV

A PDF has no table objects, only glyphs at coordinates, so the grid has to be inferred. Text sharing a baseline becomes a row. A horizontal gap wider than the tolerance breaks one cell from the next, and the left edges of those cells are then clustered across the whole page into columns. Ruling lines are ignored entirely: alignment is the only signal there is. Widen or narrow the Column gap setting first, because that single control fixes most of the damage.

Excel follows your regional list separator, so a comma file collapses into a single column on a machine set up for semicolons, switch the Delimiter option. The UTF-8 BOM toggle covers the second problem: without that byte-order mark Excel guesses a legacy encoding and turns é into é.

Columns are clustered per page, so pages that align differently can shift in the combined CSV.

Every row needs positioned text. A scanned page is a picture with no text layer under it, so there is nothing to align and the run stops with an error, send it through OCR PDF first. The other cause is Skip prose lines, which discards any line that produced a single column; turn it off if your table has a column with one value in it.