PDF.FJ.CNGuide · 5 min read · Updated 2026-08-09
How to Convert PDF to CSV
Getting a table out of a PDF and into a spreadsheet is one of the most common — and most fiddly — PDF jobs there is. Here is how to do it, and how to tell whether the result can be trusted.
This file never leaves your browser.
Convert your file
Drop a PDF containing a table below. You will get a .csv that opens directly in Excel, Numbers, Google Sheets or pandas.
How columns are detected
This is the part worth understanding, because it explains every quirk you will run into. PDFs do not mark up tables. There is no <table> tag, no cell boundaries, no column metadata — a table in a PDF is just text positioned to look like a grid, often with some lines drawn around it.
So the converter works backwards from position. It reads every word with its horizontal coordinate, groups words that share a vertical position into a row, then clusters the horizontal coordinates across all rows to work out where the column boundaries must be. Words falling in the same cluster become the same column.
Steps
- 1Choose the PDFEverything stays in your browser tab.
- 2ConvertRows are detected page by page and column positions are clustered across the whole document.
- 3Check the row and column countThe result line reports something like "12 rows × 4 cols". If the column count is wrong, you will see it here before opening the file.
- 4Download and open in a spreadsheetValues containing commas or quotes are escaped properly, so the CSV imports cleanly.
Reading the result count
The reported column count is your cheapest sanity check. If your table visibly has five columns and the tool says three, two pairs of columns were merged because they sit too close together. If it says eight, a column was split — usually because one cell contains text that wraps or is centred rather than aligned.
- Too few columns: values from adjacent columns end up in one cell, separated by a space.
- Too many columns: one logical column is spread across several output columns, leaving blanks.
- Right count but shifted rows: the table probably has merged header cells spanning multiple columns.
Common problems
| Symptom | Why | Fix |
|---|---|---|
| Paragraph text became rows of nonsense | Prose was mistaken for a table | Use PDF to Text for prose — CSV only makes sense for tabular content |
| "No tables found" | The page has no rows with repeating alignment | Confirm the content really is a table; if it is a scan, no tool without OCR can read it |
| Numbers became text in Excel | Locale decimal separator mismatch | Import with the correct locale, or use Data → From Text with explicit column types |
| Multi-line cells split into rows | A wrapped cell is genuinely two lines of text | Merge them after import — the row split is real in the source layout |
Frequently Asked Questions
Can it extract multiple tables from one PDF?
Rows are collected across the document, so several tables with the same shape merge nicely, while tables with different column counts will fight each other. For very mixed documents, split the PDF first and convert the relevant pages.
Does it work on scanned tables?
No. A scan is an image, so there are no word coordinates to cluster. OCR is required first, and this site does not do OCR.
Will merged cells survive?
Not faithfully. CSV has no concept of a merged cell, so a header spanning three columns lands in the leftmost one and the others come out empty.
Is CSV or Markdown better for tables?
CSV if the destination is a spreadsheet or a data pipeline. Markdown if the destination is documentation or an LLM prompt, since it keeps the table readable inline.