Compare PDF Files

Drop both files into the tool above and press Compare. There is no account, no page limit and no watermark, and neither file is uploaded — the comparison runs entirely in your browser.

Both, and you choose which. Text only compares the words, which is what you want for contracts and drafts where the wording is the point. Visual compares the rendered pixels and catches what words cannot: a changed signature, a moved stamp, a swapped logo, a redrawn table, a paragraph that is now bold. Text + visual runs the two together. The reason they are separate is that re-exporting a document from different software can move thousands of pixels while changing not one word, so visual comparison needs its sensitivity control to stay useful — and text comparison is immune to that noise entirely.

Each page is split into words and the two versions are matched with a longest-common-subsequence diff — the same technique used to compare source code. Working in words rather than lines matters: a paragraph that reflowed because one word was added has completely different line breaks but almost identical words, and a line-based comparison would report the whole paragraph as changed.

Yes. Side by side puts the original in the left column and the revision in the right, with each change occupying one aligned row so the columns cannot drift apart. Inline instead keeps every change inside the sentence it belongs to, which is better for judging whether an edit matters. Pages shows both versions drawn as they will print. You can switch between the three at any time without comparing again.

Both versions of a page are drawn and compared pixel by pixel on perceived brightness rather than raw colour, which is what lets it ignore the faint edge differences that anti-aliasing produces on every re-render. Changed pixels are grouped into regions and boxed, so you get a handful of marked areas rather than a fog of speckles. Highlight marks the changes on the revised page; Overlay lays the two on top of each other so you can see what moved. The sensitivity slider sets how different two pixels must be before they count.

Yes. Capitalisation, punctuation and numbers can each be ignored, and the comparison re-runs the moment you change one. Ignoring punctuation is the most useful of the three: a comma moving is the difference nobody is looking for, and it drags the words on either side of it into the result. These settings only affect what counts as a change — the text shown on screen is always what the document actually says.

It is the share of words the two documents have in common, counted against the length of both — the same measure diff tools have used for decades. Two files with no shared wording score zero; identical files score a hundred. It is a quick sense of scale, useful for telling a tidy-up apart from a rewrite before you read a single change.

It is detected and reported as such. Pages are matched on their content rather than their number, so inserting a page near the front does not shift everything after it into looking rewritten — the later pages still compare against their real counterparts. A page whose text matches one elsewhere in the other document is reported as moved, with both page numbers. A page similar to but not the same as its counterpart is reported as changed and diffed; one that resembles nothing is reported as added or removed. Page structure mode lays all of this out page by page.

For text, only if the scan has been through OCR and has a real text layer — a plain scan is a picture of a page with no text in it, so there is nothing for the word comparison to read and both sides will look empty. Visual mode has no such limitation and is the right choice here: it compares the pages as images, which is exactly what a scan is. Comparing two scans for a changed signature or an added stamp is one of the things it does best.

No. Both PDFs are read and compared inside your browser, and nothing is transmitted. That is what makes this safe for contracts, agreements, statements and drafts under NDA. Once the page has loaded it works offline.

No fixed limit. Pages are read one at a time and released as they go, so long documents are a matter of patience rather than a wall. A page with an extreme amount of text is reported as wholly rewritten rather than diffed word by word, which keeps a pathological file from locking up the tab.

Two ways. Save report writes a self-contained HTML file with the changes still highlighted, which opens in any browser and can be filed or forwarded to someone who was not at the comparison. Save .txt writes the same differences as plain text, numbered, for pasting into an email or a ticket. Neither needs this site to open afterwards.

Not directly. An encrypted file cannot be read without its password. Open it in a reader that has the password, save an unprotected copy, and compare that copy.