PDFGym
EN

Compare two PDF files

Earlier PDF or drop it here
Later PDF or drop it here
Choose an earlier PDF and a later PDF first.

Two versions of one document, and no changelog: this page finds what moved. It pairs up matching pages across insertions and deletions, then checks every pair twice, once in the selectable text and once in the rendered ink. Both files stay on your device, and the finished report is verified before you can download it.

The supplier contract with no changelog

When the other side swears nothing changed, believe the page map instead. Load both versions and let the tool point at the exact pages.

A procurement officer receives supplier-contract-v4.pdf, 28 pages, with a note saying only the formatting changed since v3. She loads v3 as the earlier PDF and v4 as the later one. The summary counts two pages needing review: page 9, where the liability cap moved from 50,000 to 25,000, and a brand-new page 22 with an arbitration clause. Formatting indeed.

Each row opens both renders side by side with the added and removed words listed beneath, so the reply email quotes chapter and verse. Ten minutes of checking against an afternoon of line-by-line reading.

Four passes from two files to one verdict

  1. Load both versions

    Press Choose earlier PDF and Choose later PDF, or drop each file into its marked area. Order matters: added and removed words are counted from the earlier file toward the later one.

  2. Compare the versions

    Press Compare PDFs. The tool reads the selectable text of both files, pairs up matching pages across insertions and deletions, then renders every aligned pair before counting changed, inserted, removed, and unchanged pages.

  3. Walk the page map

    Select any row to see both page renders with the differing regions marked, plus the Words added and Words removed samples beneath. The word check ignores capitalization and repeated whitespace, so reflowed lines do not count as edits.

  4. Download the report

    Press Download comparison report for the standalone pdf-comparison-report.html file. It is read back and row-checked before the button appears, so every aligned page in the comparison lands in the archive.

How to read the page map

The summary opens with the only number that matters: how many pages need review. Beneath it sit the four buckets, Changed, Inserted, Removed, and Unchanged, plus a running total of words added and removed across the whole pair. The proof line records how many earlier pages aligned with how many later ones, how many pairs got rendered, and how many report rows were read back.

Each row title names both sides, such as Changed: earlier page 9 · later page 9, and the captions under the previews say plainly when one side has No earlier page or No later page. Start with the changed rows, glance at the inserted and removed ones, and give the unchanged rows a skim: the fastest review reads the map first and opens the source files only where it points.

What the comparison can mislead you about

A paragraph that slid down one line paints a changed region across both its old and new positions, even though not a word differs. Big regions flag layout movement, not big edits; the word samples underneath tell you whether the text itself moved. Scanned pages get pixels only, with no word list, since the tool does no OCR: visual changes surface, silent rewordings inside a fresh scan cannot.

The report is a page map, not a legal redline. When signatures, form entries, annotations, or print-production details decide the matter, verify those in the source files themselves. When you know what moved and want to fix your copy, How to edit a PDF covers what a browser tool can change, and Edit text in a PDF explains why replacing a sentence in place is harder than it looks.

Why every page gets checked twice

Text alone lies in two directions. A scanned page carries no selectable words at all, so a text-only diff would call any two scans identical. A reprinted page can carry identical words in a shifted layout, which a text-only diff would wave through while the signature line moved half a centimetre.

That is why each aligned pair faces both checks. The word diff counts added and removed terms in the selectable layer; the pixel pass renders both pages and marks where the ink moved. A page that reads the same but prints differently, or prints the same but reads differently, lands in the review pile either way. Nothing is modified on either side: comparing only reads, and both files stay where they loaded. To diff a version as plain text with your own tools instead, PDF to Markdown turns either file into checkable text.

Three other ways to diff two PDFs

Adobe Acrobat Pro ships a Compare Files view that lists text, image, and formatting changes side by side; reach for it when the whole office already lives in Acrobat. Microsoft Word can do the job without Acrobat: open each PDF in Word to convert it to editable text, then run Compare from the Review tab. That route spots rewording well but sometimes blames the conversion for shifts Word itself introduced.

The third route is plain text. PDF to Markdown converts each version to text you can diff in any programmer editor or version control. Pick this page when the files must never leave the device, when layout matters as much as wording, or when you want the verdict in one sitting instead of a conversion round trip.

Short answers about PDF comparison

I loaded the versions the wrong way round. Does it matter?
Yes, swap them. Added words count from the earlier file toward the later one, so reversed inputs flip every added word into a removed one. The page pairs stay the same, but the story reads backwards.
Why is a page marked changed when the words match?
The ink moved even though the text did not: a shifted layout, a new stamp, a re-rendered font. Matching words pass the text check, then the pixel pass overrules when the renders differ.
Can it compare two phone scans of one contract?
It can, from pixels alone, but expect noise. Two separate scans never align perfectly, so minor skew reads as change. This works best on two exports of one digital file.
How do I compare three drafts, not two?
Run two sittings: v1 against v2, then v2 against v3. Each report names its pair, so the chain stays readable. For many drafts, export every version with PDF to Markdown and diff the text files instead.
Will a 500-page manual break the tool?
No fixed ceiling stands in the way. Everything renders on your device, so memory sets the practical limit. Heavy files simply take longer; they are not turned away.

The rest of the local toolbox

Every tool below runs the same way: your files stay on the device, each result gets checked before the download appears, and no account is asked for.

Work with the differences