PaperDiff

Compare two Word documents without opening Word

Word has a comparison feature, and it is good. It also requires Word, requires both documents to be openable on the same machine, and produces a third document you then have to read. If you have been sent a .docx on a computer without Office, or you want the answer rather than another file, neither of those is convenient.

This reads both .docx files directly in the browser and shows the differences word by word. No Office install, no account, and the documents stay on your machine, which matters more here than for most file types, because a .docx is usually a contract, an offer or an agreement in draft.

Originalthe earlier draft

or drop it here

.pdf · .docx · .txt · .md

Revisedthe one to check

or drop it here

.pdf · .docx · .txt · .md

Nothing is uploaded. Both files are read inside this tab.

How to do it

  1. 01

    Drop both .docx files

    Straight from a download folder or an email attachment. Nothing is installed and nothing is uploaded.

  2. 02

    Check how tracked changes were read

    A document with unaccepted tracked changes is read as if they had all been accepted: insertions kept, deletions dropped. That is the version the author is proposing, and it is what "the new draft" means. See below for why it is not the only defensible reading.

  3. 03

    Step through the changes

    J and K move between them. Headings, numbered clauses and table cells each keep their own place, because a .docx states its paragraph structure outright rather than leaving it to be inferred.

  4. 04

    Hand the result on

    Copy the summary for an email, export the change list as CSV, or print it. There is no page view for a Word file, for the reason set out below.

What is actually inside a .docx, and why it compares more cleanly than a PDF

A .docx is a ZIP archive of XML. The body text lives in exactly one member of it, word/document.xml, and that single fact does more for a comparison than any amount of cleverness elsewhere.

Headers and footers are in different members (word/header1.xml, word/footer1.xml), so they never reach the body text at all. The page-furniture detection that a PDF needs, and the whole class of noise it exists to remove, simply does not arise: there are no page numbers in the text to renumber. And paragraph structure is stated rather than inferred. Every w:p element is a paragraph, by definition, so a heading is a heading and a table cell is a cell, with none of the coordinate guesswork a PDF requires.

The thing a .docx does NOT contain is pages. Where Word breaks them is decided when the document is rendered, from the fonts installed on the machine doing the rendering and sometimes from the printer driver; the file does not record it. So there is no page view for a Word document, and the change list shows no page number at all rather than an invented one. Drawing an invented pagination would be the worst kind of wrong: confidently specific about a fact that does not exist.

Tracked changes are the one real editorial decision. A document with unaccepted changes contains both readings at once: insertions wrapped in w:ins, and deletions in w:del carrying their text in w:delText rather than in ordinary runs. This reader takes the document as if every change had been accepted, because that is the version being proposed and the one a reader means by "the new draft". The alternative would be comparing against a version nobody is proposing. Field codes get similar treatment: the instruction behind a table of contents or a cross-reference is skipped, because it changes whenever the document is edited anywhere, while the field’s RESULT, the text actually on the page, is compared like any other words.

Questions

Are my documents uploaded?
No. Both files are read inside the browser tab and never sent anywhere; there is no upload endpoint on this site, so it could not receive them. You can confirm it in your browser’s network inspector: no document data is transmitted.
Do I need Microsoft Word installed?
No. Nothing is installed at all. The .docx is unzipped and its XML read inside the browser tab, which is also why it works on a Chromebook, a locked-down work laptop or a borrowed machine.
What happens to tracked changes?
They are read as if accepted: inserted text is kept and deleted text is dropped. That gives you the document the author is proposing. If you need to see the tracked changes themselves, Word’s own review pane is the right tool: this compares two versions, it does not display one version’s revision marks.
Are comments compared?
No. Comments live in a separate part of the archive (`word/comments.xml`) and are not read. Neither are footnotes. Only the body text is compared, which includes tables and the text inside equations.
Why is there no page number on each change?
Because a .docx does not record where its pages break: that is decided at print time by the machine printing it. Rather than invent a page number, the change list leaves it blank and the exported report says "n/a". Compare the PDF exports of the two documents if page numbers are what you need.
Can it open an old .doc file?
No. A .doc is an OLE compound file, a completely different format from the 2007-onwards .docx, and it is detected and reported as such rather than failing with a parse error. Open it in Word or LibreOffice and save as .docx.
Is it really free?
Yes. No account, no page limit, no watermark and no trial clock. The comparison runs on your own machine, which is why it costs nothing to offer: there is no server doing the work.