PDF to Excel Converter — Free Table Extraction (.xlsx), No Upload

There is no table in a PDF

Only text at coordinates that a human eye reads as a table. There are no rows, no columns and no cells recorded anywhere; the grid you see is either drawn lines or, very often, nothing but alignment.

So converting to a spreadsheet means inferring where the columns are from where the text sits — clustering by starting position, with the tolerance scaled to the font size so a slightly ragged column still reads as one column.

Which makes every extracted table a reconstruction, and that has a consequence the rest of this page is about.

Where the inference breaks

• Merged or spanning cells have no representation to recover, so they cannot be split back out.
• A cell whose text wraps onto a second line can be read as a second row.
• Right-aligned numbers beside left-aligned labels make the column boundary genuinely ambiguous.
• Blank cells are the dangerous one. A gap where a value should be can shift every subsequent value one column left — and the result still looks like a perfectly valid table.
• Multi-line headers, footnote rows and totals rows are not marked as different from data, so they arrive as data.
• Tables continuing across a page break repeat their header, which then lands in the middle of the rows.

Therefore: verify, do not trust

That blank-cell failure is why this page insists on checking. A shifted column does not look broken. Every cell holds a plausible value of the right type, the row count is right, and nothing signals an error — you find out later, from a total that is wrong for no visible reason.

Two cheap checks catch it:

1. Sample rows containing blanks. Pick a few and read them against the original PDF. Rows with gaps are where shifts start, so that is where to look rather than at the first ten rows.
2. Total a numeric column and compare it with the total the PDF states. This is the strong check — a single misplaced value changes the sum, and eyeballing never catches what arithmetic does immediately.

Scanned documents

A scan holds a picture of a table, not text, so there is nothing to extract and no OCR is performed here. The test is the same as for any PDF: try to select the text in a viewer. If you cannot select it, no extractor will find it.

When this suits the job

It works well on a clean, machine-generated table with consistent alignment — which describes most bank statements, transaction exports and generated reports, and those are the files people most often need in a spreadsheet.

It works badly on a densely formatted document full of merged cells. And the honest advice for a small table of that kind is that retyping twenty rows is often faster than finding and fixing a misparsed column, especially once you factor in the risk of not spotting the error at all.

Everything runs in your browser and nothing is uploaded — which matters here, because the PDFs people convert to spreadsheets are overwhelmingly bank statements, invoices and payroll reports. For the words without any table inference, see PDF to text.