Finding what changed between two PDFs

What a comparison actually does

There is no change history inside a PDF. A comparison tool is given two finished documents and has to work out the relationship between them from scratch, which is a different problem from tracked changes in a word processor and a harder one.

Two approaches, and every tool uses one or both:

• Text comparison extracts the words from both files and lines them up, reporting insertions and deletions. It finds meaning changes and ignores layout entirely.
• Visual comparison renders both pages to images and looks for pixels that differ. It finds everything that moved, including a paragraph that shifted because a word above it got longer.

Compare PDF puts the two side by side and marks what differs.

Why the two disagree

They answer different questions, and the difference is not a detail.

Change one word early in a long document and a text comparison reports one change. A visual comparison may report every page after it as different, because the extra characters pushed every subsequent line along. Both are telling the truth.

The other way round: replace a logo with a different logo of the same size, and a visual comparison sees it immediately while a text comparison sees nothing at all, because no text changed.

So choose by the question you actually have:

• “Did the terms change?” Text. Use PDF to text on both files and compare the results in any text tool if you want to see it raw.
• “Is this the same document they sent me?” Visual.
• “What did they change in the contract?” Text first, then visual on the pages it flags.

What comparison cannot see

Scanned pages have no text. A text comparison of two scans finds nothing, because there is nothing to find — run OCR first, and then remember that you are comparing two sets of recognition results, so a reported difference may be an OCR error rather than a real change. Is this PDF scanned or is it text? covers how to tell.

Reading order is inferred. In a multi-column layout, extracted text may not follow the order a human reads, and two files whose columns were built differently can look different when they are not.

Invisible changes stay invisible. Metadata, embedded attachments, revision history and anything hidden behind a covering rectangle are not compared by a tool looking at words or pixels. If what matters is whether a file was tampered with, comparison is the wrong instrument — a signature is the right one. See how to sign a PDF.

Identical-looking is not identical. Two files can render the same and differ completely in bytes: different producers, different fonts embedded, different compression. Nothing is wrong when that happens.

A workflow that works

1. Check the obvious first. Page counts and file sizes, via Page count and file info. A different page count tells you where to look before any comparison runs.
2. Text-compare the whole document to find where the wording changed.
3. Visually compare the pages it flagged, plus any page that matters regardless — the signature page, the schedule of figures.
4. Pull out the disputed pages with Extract pages if you need to send someone a specific section rather than the whole file.

For long documents, comparing sections is far more practical than comparing two-hundred-page files in one pass — and it makes the result something a person can actually check.