Compare PDFs

1. 01Put the old version in one field and the new version in the other. On a phone, the two fields are stacked; on a computer, they are side by side.
2. 02Tap “Compare”. The text is taken from both files page by page. For thick documents this takes a moment.
3. 03At the top you see how much has changed and on which pages. Tapping “Page 5” jumps to the spot.
4. 04How to read it: − is the old passage, + the new one. The changed words are bold and underlined. The two numbers on the far left are line numbers within the page, first the old version, then the new one. They start again at 1 on every page. With “Join line breaks”, they show the first line of the paragraph.

The text comes from the text layer of the PDF, just as it would when you select and copy it. The pieces a PDF stores are put back together into lines: grouped by height, sorted from left to right, spaces derived from the gaps, hyphenation at the end of a line resolved. The comparison then uses the Myers algorithm, the same one as in Compare texts.

Where it reaches its limits: With multi-column layouts (newspaper style), the columns are read across, line by line, so the order is wrong. Text that runs sideways on the page does not reliably land in the right line. Tables become lines with spaces in between. The tool does not detect columns. Headers and footers count as normal text: if they say “Page 3 of 12” and the new version has one page more, every page counts as changed. And if a PDF has no text layer at all (scanned), it stays empty here.

“Join line breaks” is on by default on purpose: two versions of the same contract almost never break lines at the same place. Without this setting, a single inserted word turns a whole paragraph into a difference. To find a paragraph, however, the tool has to guess: from line spacing, line length and font size. That can go wrong, and then a piece of unchanged text appears as removed and right below it again as added. If a paragraph runs across a page break, it is only joined again when the continuation is clear (a hyphen at the end of the page, or it continues in lower case). Otherwise both halves stay separate and both count as changed. If you need it exact, turn the setting off: line by line nothing is guessed, and the comparison runs line by line just as the lines are in the file.