Two ways a table is found
By grid lines: when the PDF contains borders and rectangles, the structure is rebuilt from them. Invoices, acts, price lists, reports and exports from ERP systems or Excel are almost always drawn this way — the card labels such a table “Found by grid lines”.
By text alignment: with no borders, columns are detected from the gaps between fragments on the same line. That is how “invisible” tables are found — specifications, item lists and price lists without a grid.
- Grid lines are the most accurate case: merged cells and borders are preserved.
- Text alignment works without any borders, but the result deserves a comparison with the original.
- The method is printed on the card, so you know how much to trust the structure.
A table that runs over several pages
When a page break cuts a table in two, “⤒ Merge with previous” joins the fragments. A long price list or register becomes one table instead of ten pieces with a repeated header.
When the structure is not quite right
The result can be fixed before exporting — faster than redoing the file in Excel:
- Double-click edits the text of a cell.
- “Header rows” and “Drop columns on the left” decide what goes into the file.
- A block that is not a table is removed with the cross; a single table can be downloaded on its own.
Frequently asked questions
Can I extract just one table from a document?
Yes. Every table has its own card with “Excel”, “Word” and “Copy” buttons, so the other tables are not downloaded with it.
What if the PDF has no visible borders?
Nothing to configure: the service switches to column detection by text alignment on its own. The card then says “Found by text alignment”, and that result is worth comparing with the original.
Does table extraction work on scans?
No. A scan has no text layer, and there is no OCR in this service. Run the document through Adobe Acrobat, ABBYY FineReader or another OCR tool first, then upload the resulting PDF — its tables will be extracted normally.