See what changed between two PDFs

Compare the wording and the appearance of two versions of a document — quotations, reports, contracts — without uploading either of them. Both files are read in this browser and never leave your device.

Home › PDF Tools › PDF Compare
Processed on this device · nothing is uploaded · nothing is saved
What that means

Both PDFs are read by code running in this browser tab. There is no upload, no server and no analytics on this address, and the PDF viewer it uses is served from here rather than from someone else's network. Nothing is stored unless you download a report yourself.

Step one

Choose the two versions

The sample is an invented quotation, before and after.

What this tool does, and what it does not

How the comparison works
  • Pages are paired before they are compared. Inserting one page near the front of a document would otherwise shift everything and make every page after it look rewritten. Pages are matched by the words they share, allowing for insertions and deletions, and pages with no partner are reported as added or removed.
  • Nothing meaningful is normalised away. Spacing and line breaks are ignored; digits, minus signs, decimal points and currency symbols never are. Ignoring case or rejoining hyphenated words is available, off by default, and recorded in the report.
  • Text and appearance are counted separately. An edited line is usually both a text difference and a visible one, so adding them together would count it twice while sounding more precise.
  • “Contains digits” means exactly that. It is a filter for finding numbers quickly, not a judgement about money or importance.
  • Nothing here interprets your documents. There is no summary of what a change means, and no opinion on whether anything is safe to sign.
Limits, honestly
  • A PDF stores text as pieces at coordinates, not as sentences. Reading order is reconstructed from where those pieces sit, and on complicated layouts it can be wrong. Pages where the text could not be read confidently are marked, and the extracted text is available to check.
  • Scanned pages have no text to compare. They are compared as pictures instead, and the report says so rather than calling it a text comparison. There is no text recognition here.
  • The visual comparison has two thresholds — how different a pixel must be, and how large a changed area must be — and both can hide a small change. They are shown in the report, and a strict setting is available.
  • Pages of different sizes are compared at their actual size rather than stretched to match, because stretching one to fit the other makes every line look moved.
  • Finding no differences is reported as “no text differences found in the pages compared”, with the coverage. Only a byte-for-byte check can say two files are identical, and that is reported separately when it applies.
  • Both files are held in memory. Comfortable up to roughly 100 pages each; beyond that, compare a range of pages at a time.
  • Nothing is saved between visits — download the report before closing the tab.

Comparing plain text or prose rather than PDFs? Text Compare does that. Comparing two spreadsheets? Spreadsheet Cleanup & Reconciliation matches rows by a key column.