PDF

Convert a PDF table into an Excel spreadsheet (.xlsx)

Detects tables in a PDF by column alignment and row regularity, shows a preview before generating — with adjustable column boundaries and the option to discard a table detected by mistake — and produces an .xlsx with one sheet per table or per page, your choice.

Click to choose a PDF

Up to 50 MB, no password protection

Local processing: your file never leaves your computer.

Does:Detects tables in a PDF by column alignment and row regularity, shows a preview before generating — with adjustable column boundaries and the option to discard a table detected by mistake — and produces an .xlsx with one sheet per table or per page, your choice.
Processes:Your PDF is processed entirely in your browser — nothing is uploaded to a server.

The problem with "reading" a table inside a PDF

A PDF doesn't store a table as a table — it stores positioned text. What makes a grid of numbers look like a table to a human is purely visual alignment: each column always starts at the same horizontal position, row after row. This tool uses exactly that signal — alignment that repeats — to recognize where a table probably is, without relying on any explicit table markup, because the PDF simply doesn't have one.

How the detection works

First, the horizontal starting position of every line of text on the page is recorded; positions that repeat across several lines become candidate column starts. Then, runs of at least two consecutive lines whose text falls at two or more of those recurring positions are treated as a table. The two conditions together — column alignment and regularity across several lines — cut down the chance of mistaking ordinary text (which rarely holds that pattern for more than one line) for a real table.

Why the preview comes before the final file

Automatic alignment-based detection is an estimate — it works well for a table with a regular grid, but it can get ambiguous cases wrong: a table of contents with a title on one side and a page number on the other has exactly the pattern of two aligned columns, without being a real data table. That's why the result shows up first as an editable preview, before any file is generated — you can check whether each identified table actually makes sense before downloading anything.

Adjusting a column boundary

Every detected table shows the list of horizontal positions used to separate its columns, as editable numbers. When a column comes out incorrectly split into two, or two columns get merged into one, that list can be edited directly — the preview regroups each row's text immediately using the new positions, with no need to reprocess the PDF. It's a manual adjustment on top of an automatic result, not a second detection attempt.

Discarding a table detected by mistake

Not every run of aligned lines is actually a data table — the table of contents example above, or a block of text with regular indentation that happens to line up at more than one position. Any detected table can be discarded individually before generating the final file, without affecting the others — the generated file only includes the tables you kept.

One sheet per table, or one per page

There are two ways to organize the final result: one Excel sheet for each kept table (useful when each table is an independent dataset, even if they come from the same page), or one sheet per page, combining all the kept tables from that page, separated by a blank row — useful when several small tables on the same page are part of the same context and make more sense viewed together.

Numbers, currency, and dates come out as the text that was read, not recalculated

Every cell in the generated spreadsheet gets exactly the text read from that position in the original table — a value like "$1,234.56" or "12/31/2026" comes out as that literal text, without being converted into a number or a date Excel recognizes automatically. This faithfully preserves what was written in the PDF, but it may take an extra step inside Excel (using "Text to Columns" or reformatting the cell) if the goal is to sum or calculate on top of those values later.

Why only the text goes into the preview, with no visual styling

The preview keeps column alignment, but doesn't reproduce background color, cell borders, or bold text the original table might have had in the PDF — the focus of detection is finding where the data is and how it's organized into rows and columns, not replicating the table's visual appearance. The generated .xlsx file follows the same logic: organized data, without the source table's visual formatting.

What this tool doesn't do

It doesn't read a table inside an image — only text that already exists as text in the PDF; for a photographed or scanned table, this site's OCR tool recognizes the text first. It doesn't recover formulas: a PDF doesn't store spreadsheet formulas, only the already-calculated result, so every cell in the generated file comes out with a value, never a calculation. And it doesn't recognize a table without consistent column alignment — a PDF that visually represents tabular data but without real aligned columns behind it (irregular manual spacing, for example) may not be detected.

Frequently asked questions

How does the tool recognize a table inside the PDF?

It looks for runs of at least two consecutive lines whose text lines up at two or more horizontal positions that repeat several times on the page — the same visual pattern as a real table (aligned columns, line after line). Ordinary running text, without that repeated alignment, isn't read as a table.

Why does the preview show up before the file is generated, instead of just the final result?

Because automatic table detection by alignment is an estimate — it can correctly find a real table, but it can also mistake column-based text (like an index or a two-column layout) for a table. The preview lets you check that before spending time generating and downloading the file — and lets you adjust the column boundaries or discard a table detected by mistake before generating anything.

How do I fix a column boundary that came out wrong?

Each detected table shows the list of horizontal positions used as column boundaries, editable as text (numbers separated by commas, in points). Changing these values regroups each row's text immediately in the preview, with no need to reprocess the whole PDF.

Does the tool read a table inside an image, or recover the original spreadsheet's formulas?

No, neither. It only recognizes tables from text that already exists as text in the PDF — a table photographed or scanned as an image isn't read (for that, this site has an OCR tool). And since a PDF doesn't store spreadsheet formulas, only the already-calculated result as text, no formula is ever recovered — the cell comes out with just the value.

What's the difference between one sheet per table and one sheet per page?

One sheet per table creates a separate spreadsheet for each table that's detected and kept, useful when each table represents an independent dataset. One sheet per page puts all the tables kept from the same page into the same spreadsheet, separated by a blank row — useful when several small tables on the same page are part of the same context.