How to Convert a PDF Table to Excel (XLSX or CSV)

Text-based PDFs only

PDF to Excel reads the embedded text layer of a PDF — the actual words a PDF editor put on the page. Documents created by Word, Excel, a browser's Print to PDF, or invoice and report software have such a text layer. Scanned or photographed pages do not: they are pictures, so there is nothing to read.

If you select a scanned PDF, the tool tells you and points to PDF OCR. Run OCR there first, then convert the recognized text file to spreadsheets here, or simply reuse the OCR tool's .txt output.

How the columns are detected

The tool groups the page's words into lines by their vertical position, using the exact text coordinates the PDF stores rather than the visible table rules. Because gridlines themselves are ignored, it is this positioning that defines the table.

A large horizontal gap between words — more than about 24 points, roughly a third of an inch — starts a new column. Uniform tables with clearly separated columns convert cleanly. Columns that butt against each other, spanning cells, or rows that merge across columns come out differently than the visual layout and usually need a quick cleanup in Excel.

Download, then check the data

Every extracted cell is exported as text: a cell that looks like a number is a label, not a numeric value, so sums and charts may need a one-step conversion in Excel (select the column, then convert text to numbers). This is deliberate — values are never interpreted as formulas, so a pasted dashboard formula from the PDF cannot run unexpectedly on open.

Preview the grid on the page first; the preview shows the first 200 rows. Large documents are capped at 100 MB and 200 pages, keeping the conversion fast and the tab responsive. Nothing is uploaded — the whole conversion runs in your browser.

When the extraction looks wrong

Two common causes: the PDF uses a multi-column page layout where side-by-side columns interleave into one line, or the source PDF was generated in a way that made its text coordinates unreliable. For tightly designed layouts, consider running OCR first for a flatter text stream, then converting that text here.