PDFJadoo › Blog › PDF Tables to CSV

PDF to CSV: rows and columns, liberated from the page

Data in a PDF is data in a display case: visible, untouchable. The register, the price list, the results table — to sort it, sum it or join it with other data, it must become rows and columns again. PDF to CSV performs the extraction geometrically: one row per text line, columns split where the gaps are — the white-space channels that separate a table's columns on the page become the CSV's commas — with an optional page-number column so every row stays traceable to the page it came from. The limits are stated on the tin: it works on text-based PDFs, it is not layout-perfect (merged and multi-line cells need a tidy), and scans need OCR first. This article explains the geometry, the traceability trick, and the tidy that honest extraction expects.

Free, in your browserFiles never uploadedRows from lines · columns from gaps
PDFJadoo logoBy the PDFJadoo Team5 min read
Diagram: a PDF table's rows flowing out into a comma-separated file
The gaps between columns were always commas waiting to happen.

How geometric extraction works — and where it bends

A born-digital table is text placed at coordinates. The extractor reads the text line by line — each line a row — and looks for the consistent vertical gaps: the channels of white that run down the table between columns. Where the gap pattern says a column boundary lives, a comma is written. For the great run of real tables — registers, statements, lists, results — the geometry is regular and the extraction is faithful. The bends are equally geometric: a merged cell (one value spanning two columns) has no gap inside it, so it lands in one column and shifts its row's alignment; a multi-line cell (a description wrapped over three lines) becomes three rows, because lines are rows. Neither is a defect of effort; both are what “not layout-perfect” honestly means — the CSV is a faithful transcript of the page's lines, and the spreadsheet tidy (merge the wrapped rows, fix the spanned cells) is the expected second step, sized in minutes for real tables.

The page-number column is the professional's option: with it on, every row carries its source page. When a figure is questioned three weeks later — “where did this total come from?” — the row answers: page 14. Extraction without provenance is data; extraction with provenance is evidence. For anything destined for audit, reconciliation or dispute, tick the column.

Choosing the right extraction for the data

Three siblings divide the work: this tool for CSV (the universal data format — every system imports it), PDF to Excel when the destination is a spreadsheet with formulas and types, and plain text when the content is prose rather than data. And the precondition governs all three: the PDF must carry text. A scanned table is a photograph of data — OCR it first, then extract.

Extract a PDF table to CSV, step by step

Step 1 — confirm text. Select a cell's text in the PDF: selectable means extractable. A scan goes to OCR first.

Step 2 — drop the PDF onto PDF to CSV. Extraction runs in your browser — registers and statements stay local.

Step 3 — decide the page-number column. On for anything auditable; off for a clean data file whose source is obvious.

Step 4 — press Extract to CSV and open the result in a spreadsheet. Check the known trouble spots: the merged header, the wrapped description, the totals row.

Step 5 — tidy, then trust. Fix the spanned and wrapped rows the geometry warned about; spot-check three figures against the PDF; then build on the data.

The tidy's checklist is short

Real tables tidy fast because their irregularities cluster: one merged title row, a handful of wrapped cells, a totals line that spans. The extraction's honesty about its method — lines to rows, gaps to columns — tells you exactly where to look, which is most of what “trustworthy tool” means in practice.

Data belongs in rows, not in a display case.

Lines become rows, gaps become columns, and every row can carry its page number — extraction with provenance.

Open PDF to CSV

A worked example: the supplier price list, reconciled

A supplier's 22-page price list arrives as a PDF; the purchasing system wants a CSV. Extraction with the page column on produces 1,900 rows in a minute. The tidy finds what the geometry predicted: the list's section-header rows (merged across all columns) sit in column one with empty companions — reclassified as headers and dropped in two minutes of filtering — and a dozen wrapped descriptions split across rows, rejoined by sorting on the page column and eyeballing the seams. Three spot-checked prices match the PDF exactly. The import proceeds — and when a price is disputed a month later, the page column walks the question back to page 14, row and source together.

Edge cases

Borderless, gapless tables. Columns separated by neither lines nor clear gaps (tight numeric grids) can split ambiguously — the Excel route or manual column work may serve better; judge from the first page's output.

Statements with running balances. Multi-line transaction descriptions are the classic wrap case; the page column plus sorting makes the rejoin mechanical.

Mixed content pages. Prose around the table extracts as rows too — extract, then filter to the table's row pattern in the spreadsheet.

Before you finish: the 60-second check

Fair questions

How do I extract a table from a PDF to CSV?

Drop the text-based PDF onto PDF to CSV and press Extract to CSV. Each text line becomes a row and columns are split where the gaps between them are. Tick the page-number column if rows should stay traceable to their source pages.

Why is the extraction not perfect for some tables?

Because it reads lines and gaps, not the table's logical structure: a merged cell lands in one column, and a multi-line cell becomes multiple rows. The tool states this plainly — a short tidy in a spreadsheet fixes the few affected rows.

Can I extract from a scanned PDF?

Not directly — a scan has no text to extract. OCR it first (Make a Scanned PDF Searchable), then extract from the searchable result.

What is the page-number column for?

Provenance: every row carries the page it came from, so any figure can be traced back to its source during audits, reconciliations and disputes. Turn it on for evidence-grade extraction.

Should I use PDF to CSV or PDF to Excel?

CSV when the destination is a system, a database or another tool — it is the universal format. Excel when a person will work the data with formulas, types and formatting in a spreadsheet.

Is my document uploaded?

No — extraction runs in your browser; statements and registers never leave your device.

Tool behaviour — one row per text line, columns split at gaps, the optional page-number column, text-based PDFs only, the not-layout-perfect statement and the OCR-first rule for scans — is from PDFJadoo's own PDF to CSV page. Checked 18 November 2026.

Keep reading