PDF Table to CSV

PDF table to CSV

Choose a text-based PDF

Turn simple, text-based PDF tables into spreadsheet-ready CSV or TSV files without sending the PDF to a server. The browser groups text fragments by their visible positions, looks for repeated column anchors, and leaves ordinary prose out when it can. Review every detected table before downloading because this is a conservative, best-effort extractor, not a guarantee that complex PDF layouts will become perfect rows and columns.

How to extract a PDF table as CSV

  1. 1

    Choose a text-based PDF

    Select a PDF up to 20 MiB and 150 pages. Scans that contain only page images need OCR before this tool can read them.

  2. 2

    Let the browser detect tables

    The extractor uses visible text coordinates, including page rotation and the CropBox, then looks for repeated row and column positions.

  3. 3

    Review and choose the output

    Select the detected tables you want and check their padded rows. Choose comma, semicolon or tab, plus the UTF-8 marker and formula-protection options.

  4. 4

    Download each selected table

    Each selected table becomes its own local CSV or TSV download. Tables are not stitched together across pages.

What the detector can and cannot infer

PDF files do not normally store a table as a grid of named rows and columns. They store text fragments at coordinates. This tool groups fragments on the same visible line, merges nearby fragments, and keeps column positions that repeat across several rows. It can separate multiple regular tables on one page and ignores single-column prose when the pattern is clear.

Page rotation and an offset CropBox are included in the visible-coordinate calculation. For predominantly right-to-left rows, column order follows the direction reported by the PDF text layer. These checks still cannot reconstruct every layout.

PDF layout Likely result What to check
Clean export with repeated columns Usually the best case Headers, decimal marks and blank cells
Missing value in a regular row Empty cell padded into the inferred column Whether the correct column stayed empty
Merged or wrapped cell May split, shift or join text Compare the preview with the PDF
Several regular tables on one page May be detected separately Select only the tables you need
Scanned page image No table detected Run OCR first

The 20 MiB and 150-page limits are joined by text-fragment and character safety caps. A highly fragmented PDF can therefore stop even when its file size and page count are below those headline limits.

CSV, TSV and spreadsheet safety

Comma and semicolon outputs use the selected delimiter; tab output is saved as TSV. Records use CRLF line endings. A field is wrapped in double quotes when it contains the active delimiter, a double quote, a carriage return or a line feed, and embedded double quotes are doubled. This follows the common rules documented in RFC 4180, adapted to the chosen delimiter.

Formula protection is on by default. If a cell begins with =, +, - or @ after leading whitespace, including supported full-width variants, the tool prefixes an apostrophe so spreadsheet software is more likely to treat it as text. That prefix changes the cell value. Turning protection off preserves the extracted text exactly, but opening untrusted formula-like content in a spreadsheet can be dangerous. OWASP’s CSV Injection guidance explains why no mitigation works universally across every spreadsheet and later re-save.

An optional UTF-8 marker helps some spreadsheet applications recognize non-Latin text. Whatever settings you choose, inspect the downloaded file before using it for accounting, reporting or automated imports.

Private step state

The PDF, detected tables and choices remain in this browser. They are kept in bounded local browser storage for up to 30 minutes so the three steps can survive navigation, and downloads are created locally. Nothing is uploaded by this tool. Starting over removes the current job; the browser may also clear local data earlier because of storage policy or private-browsing restrictions.

Frequently Asked Questions

No. PDF reading, table detection and CSV or TSV creation happen in your browser. The current job is held in bounded local browser storage for up to 30 minutes so the steps can continue, and every download is local.

Not by itself. A scan that contains only page images has no PDF text layer for this extractor to position. Run OCR first, then use the resulting text-based PDF.

Choose comma, semicolon or tab. Records use CRLF, and fields containing the chosen delimiter, a quote, a carriage return or a line feed are quoted; a quote inside a field is doubled. Tab output uses the .tsv extension.

It prefixes an apostrophe to extracted cells that begin like spreadsheet formulas, including =, +, - and @ after leading whitespace. It is on by default and changes those cell values. Turn it off only when preserving raw text matters and you trust the PDF.

Detection depends on repeated text positions, not real PDF table cells. Wrapped text, merged cells, unusual spacing, mixed directions and decorative layouts can split or shift values. Rotation and CropBox offsets are normalized, but you should still compare every preview with the source.

It can detect more than one regular table on a page and lets you select them individually. Each selected table is downloaded as its own file. It does not join a table that continues onto another page.

Related Tools