Convert PDF to Excel, even when the PDF is a scan

Why PDF tables trip up general AI assistants

A PDF does not store a table as a table. It stores characters at positions on a page, and the rows and columns exist only for the eye. In a scanned PDF there are not even characters, only a picture of the page. Rebuilding the grid means deciding, for every number, which row and which column it belongs to.

A chat assistant can do this for a short, clean table. On a wide table, an empty cell or a merged cell can push every value after it one column to the right, and the result still looks like a tidy table. On a long document the assistant has to reproduce hundreds of rows in one reply, so it tends to shorten the output or stop. And what you get is often a table drawn in the chat window, which you then have to copy into a spreadsheet yourself.

This tool reads each page on its own and returns structured data: a list of columns and, for each row, exactly one cell per column, with empty cells left empty. It is instructed to copy values, not to interpret them, and to write [illegible] where a value cannot be read. Because the result sits beside the page image, a shifted column is easy to spot before you export.

What people convert

• Financial reports and annual accounts with pages of figures
• Price lists, catalogs and product sheets from suppliers
• Invoices and purchase orders with line items
• Scanned or printed tables that only exist on paper
• Research papers and government reports with data tables
• Timetables, rosters and inventory lists

How values and tables arrive in the spreadsheet

Values are copied the way they are printed. A plain number such as 1204.5 arrives as a number you can add up. A value printed with a currency symbol or a thousands separator, such as $1,204.50, is kept exactly like that and arrives as text, so nothing is silently changed. If you want to calculate with those cells, convert them in your spreadsheet after checking them.

Tables are kept separate. In the XLSX file each table gets its own sheet, named after its title or its page. A table that continues over several pages comes back as one table per page, in page order, and is not joined automatically. In the CSV file the tables follow one another in a single file. A page with no table but with clearly tabular content, such as label and value pairs or an itemized list, also comes back as a table.

Getting a clean result, and where the limits are

Digital PDFs and clean scans read best. For a photo, lay the page flat, shoot from directly above and fill the frame with the table. Very small print on a large page reads better if you split it into parts. Password-protected PDFs cannot be opened, so remove the password first.

The tool extracts data. It does not carry over fonts, colors, borders, formulas or charts, and it does not turn a chart into numbers. No reader is perfect, so compare totals and key figures with the page shown beside the table, and look at any cell marked [illegible]. For bank and credit card statements, use our bank statement converter instead: it returns every transaction in one table with fixed columns.