How to Compare Two PDFs and See Exactly What Changed

The contract comes back from the other side's lawyer asagreement_v3_FINAL.pdfwith no track changes and no note about what moved. Reading both versions side by side works, sort of, until you're nine pages in and no longer sure whether that clause was always worded that way. A text diff of the two files finds the edits for you in a few seconds.

Upload the original first

The order matters, because it's what decides which colour a line gets. Upload the earlier version first and the revised one second, and the report reads the way you'd expect: lines only in the first file are deletions in red, lines only in the second are additions in green, and everything untouched stays black. Get the order backwards and every addition shows up as a deletion, which is confusing rather than wrong.

What comes back is a PDF report, not a marked-up copy of your document. It opens with a one-line summary of how many lines were added and removed, then lists the full text with + and − markers down the left. You can send it to someone without also sending both source files.

It compares the words, not the look

The comparison runs on text. That's the point when you're checking a contract, a policy or a spec, and it's a limitation worth knowing about when you're checking a design:

• A retyped clause, a changed number, a deleted paragraph — all caught.
• A different font, a shifted margin, a swapped logo, a redrawn chart — none of these register, because the words are the same.

So a report saying no differences were found means the two files say the same thing, not that they look identical. If layout is what you're checking, convert both to images and flick between them instead.

Why one small edit can light up a whole paragraph

The diff works line by line. In a PDF, a line is whatever the layout engine put on that visual row, so inserting a single word near the start of a paragraph pushes every following word onto a different line. Nothing after that point matches, and the whole paragraph shows as removed and re-added.

It looks alarming and it isn't. Read the red and green blocks as a pair and the actual change is usually obvious within a sentence. Two things reduce the noise:

• Compare versions that were exported from the same source document at the same page size. Re-exporting a Word file to a different paper size rewraps everything and makes the diff useless.
• Compare the relevant chapter rather than the whole 200-page manual. Split both files to the same page range first and the report stays short enough to actually read.

Scanned files need OCR first

A scan is a picture of a page. There's no text in it to compare, so the report comes back either empty or nonsensical. Run both files through OCR first, which adds a text layer without changing how the pages look, then compare the results. Accept that OCR mistakes will show up as differences: a stray rn read as m is a real change to the extracted text even though both pages are identical.

Common questions

Can I compare a Word document against a PDF?

Convert the Word file to PDF first and compare the two PDFs. Expect some wrapping noise, since Word and whatever produced the original PDF won't break lines in exactly the same places, but the substantive edits still stand out clearly.

Does the report show me where on the page a change happened?

Not as a page reference. The report follows the document's reading order top to bottom, so changes appear in the same sequence as in the original, but you'll be searching the changed line in the source file to see it in context. For most review work that's a copy and a Ctrl+F away.