Compare Two PDFs

How to use the Compare PDFs tool

1. Add the original PDF, then the newer one. Order matters.
2. Optionally ignore capital letters or punctuation.
3. Press Compare and read the differences page by page.
4. Download the text report if you need a record of it.

How the comparison is made

The text of each page is read with its positions, assembled into lines in reading order, and then the two sets of lines are compared. Lines that match are passed over; lines that differ are paired up and compared again word by word.

That two-level approach is what makes the result readable. A line-only comparison tells you "this line changed" and leaves you squinting at two long sentences. A word-only comparison on a whole page produces a confetti of tiny changes. Doing lines first and then words inside them points straight at the word that moved.

Matching uses the standard longest-common-subsequence method, with identical beginnings and endings removed first. In practice that means a one-word edit in a forty-page contract is found almost instantly rather than after a long wait.

What it finds, and what it cannot

It finds added lines, removed lines and changed lines, with the specific words marked. That covers the questions people actually have: did the price change, did a date move, has a clause been added, is the bank account the same as last time.

It does not find formatting changes. Bold, colour, font, spacing and the position of a picture are all appearance, and this tool reads words. Two documents that look quite different can compare as identical if the words are the same, which is occasionally surprising and always worth knowing.

It also does not know about tables as tables. A changed figure inside a table shows up as a changed line, which is usually enough to see what happened, but the column it sits in is not named.

Pages that do not line up

Comparison is page against page: page one with page one, page two with page two. When a page has been inserted in the middle of the newer document, everything after it shifts, and from that point on the comparison reports large differences on every page.

That is not a malfunction - it is telling you truthfully that page four of one document no longer corresponds to page four of the other. The quickest way to deal with it is to use Extract Pages from PDF on both files to line up the sections you care about, and compare those.

If one document simply has more pages than the other, the extra pages are listed as existing only on that side.

Ignoring case and punctuation

Two options exist because two kinds of noise are common. Ignoring capital letters is useful when a document has been retyped or passed through a system that changed headings to title case; without it, every heading reports as changed.

Ignoring punctuation helps where a document has been through a word processor that converted straight quotes to curly ones, or hyphens to dashes. Those changes are invisible to a reader and would otherwise flood the result.

Leave both off for a legal review, where a comma genuinely can matter, and turn them on when you are looking for substantive changes in something that has been reformatted.

Scans, and why they compare as empty

A scanned PDF contains pictures of pages. There is no text in it, so there is nothing to compare, and the tool says so rather than reporting that two scans are identical - which would be technically true and completely misleading.

Run both files through OCR PDF first. That adds a real text layer, and the comparison then works normally.

Bear in mind that OCR makes occasional mistakes, and two separate OCR runs can make different mistakes on the same word. Some of the differences you see between two scans will be recognition artefacts rather than real edits, so check anything surprising against the pages themselves.

Nothing is uploaded

Files are not uploaded - everything happens in your browser. This matters especially here, because comparing documents means two versions of a contract, a tender, a settlement or a policy - exactly the material nobody should be posting to an unknown server.

Both files are read locally and the comparison runs in this tab. The downloadable report is generated in memory. Nothing is transmitted and nothing is stored.

The PDF reader is fetched once, after you choose your files, and cached by your browser afterwards.