Compare PDF shows too many changes, or misses one?

What you see, and where to look

Compare PDF reads the words of both files and compares them in order. It's a very good way to find changed words, and it has clear limits. Match what you see:

Far too many changes

The comparison puts all the words of each file in one long line and looks for the longest runs that are the same. What is left between those runs is a change. So a new sentence on page 1 pushes the text after it down, but only the new sentence is reported. When the result shows hundreds of changes, work through these causes in order:

1. Are they really two versions of the same document? If fewer than 35% of the words match, the summary says the PDFs have little text in common. Check that you picked the right two files.
2. Is "Ignore extra spaces and line breaks" on? It is on at first, and it should stay on. Then a line that breaks in a new place, or a word split by a hyphen at the end of a line, isn't a change. In our test, we laid out the same 21 pages of text again on Letter paper with other margins, so it came out as 20 pages. With the option on, the result was "No differences in the text". With it off, the same pair gave 830 changes, every one a line break in a new place.
3. Was one file made in a different program? Its text may come in another order, for example in two columns. The List view then helps you see what really changed, and the Pages view shows the real pages.
4. Was a table rebuilt? A table is read as words, usually row by row. A changed figure in a cell is found. If the table was built again, with its columns in a different order, the words come in a different order and you'll see many changes.
5. Did a paragraph move? Moved text shows as removed where it was and added where it went. That's two changes for one move.

Headers and footers have their own care. A line near the top or bottom that repeats on at least three pages (two in a PDF of up to four pages), like "Confidential" or "Page 3 of 20", is taken out of the main comparison. Page numbers that count up, and the total number of pages, are not differences. A header that really changed, say from "Version 1" to "Version 2" on every page, is reported once, as one change in a header or footer. In our test, that gave exactly one change.

If you know both files are versions of one document and the count is still huge, read the List: the first few cards usually show the pattern, such as the same kind of change on every page.

"No differences", but something changed

The word comparison looks at words and their order. It does not see things that aren't words. Check this list against what you know changed:

You can change either option after comparing. The files are compared again straight away, and Save report notes which options were used. In our test, a word changed only in its capitals gave 1 change, and none with Ignore capitals on.

One more habit: before you sign anything, do a last check with Compare the look. It shows a new stamp, a missing signature or a changed logo that the words alone don't. The post Compare contract versions and find changes before signing has a checklist.

"These PDFs have no text to compare"

A scan is a picture of a page. It has no text inside, so there are no words to compare. When one or both files are scans, press Compare the look. It draws each page of both files as a picture, 800 dots across, lines the two up (a scan can sit a few dots off), and compares them colour by colour. Areas that differ are grouped into boxes and marked on both versions, page by page.

• It works best on two scans of the same kind, such as two scans from the same scanner. A scan and a text PDF never line up exactly, so comparing their look marks most of each page.
• It can't list the words. It only shows where the pages look different. In our test, two 3-page scans with one word changed on page 2 showed that only page 2 looked different, in 0.6 seconds. A 42-page, 50 MB scan against a copy of itself showed the pages looking the same, in 18 seconds.
• Scans made with text recognition (OCR) carry hidden text under the picture, and the word comparison uses it. The result is then only as good as the recognition: a letter read wrongly in one copy shows up as a change.
• Only photos of the pages? Turn them into one PDF with JPG to PDF, then compare by look. Take the photos straight on, in the same light, so the pages match.

If you can, ask for the PDF itself, not a scan: a PDF saved from the document can be compared word by word.

The changes look backwards

"Added" means words that are only in the changed version, and "removed" means words that are only in the original. If the newer file went in as the original, it all looks the wrong way round. The fix takes one click:

• Press the swap button between the two files. The changes are found again the other way, so what was added now shows as removed.
• Or put the older file first from the start. If you pick or drop both files at once, the older file goes first by itself.
• To replace one file without starting over, use Change or Remove under it.

Strange letters in the list

Some PDFs, often ones typed with older Hindi fonts, store their letters in a code of their own. The page shows the right letters, but the text underneath is a different code, so the list shows odd signs. This is not a mistake in the comparison. The PDF is built that way.

• Use the Pages view, which shows the real pages with each change marked on them.
• Or use Compare the look, which compares the page pictures and doesn't need the text.
• Documents typed in Unicode compare well: Hindi and other scripts are compared word by word, with their vowel signs. Phone keyboards type Unicode.

It is slow, or something else won't work

Most of the time goes into reading the text of every page, and that depends on how a PDF was made as much as on its length. In our tests on a computer, two 21-page PDFs took 0.5 seconds, a 117-page report and its compressed copy took 4.3 seconds, and two 298-page PDFs took 4.6 seconds. A phone is slower. Meanwhile:

• On a phone, close other apps and tabs to give the page memory. You can press Cancel at any time: your files stay ready to compare again.
• On a phone you see one version at a time. Use the Original and Changed switch, and use List, which shows each change with the words around it.
• Copy list does nothing. Some browsers block copying. Use Save report, which saves the same list as a text file.
• "This PDF is locked." Type its password in the box that appears and press Open. It is used only on your device.
• "This file can't be read as a PDF." The file is damaged, cut short, or not really a PDF. Download it again.
• A Word file? Only PDFs can be compared. Save the Word file as PDF first, or use Word to PDF.

A routine that avoids most surprises

1. Put the older file first and the newer one second.
2. Leave Ignore extra spaces and line breaks on. Turn it off only when lines matter, as in a poem or an address block.
3. Read the summary, including the notes about little text in common or pages without text.
4. Go through the changes with the arrows, or the N and P keys, and read each as words: "1" → "15" on page 1 is a different thing from a rewritten sentence.
5. Do a last Compare the look for stamps, signatures and logos.
6. Press Save report and keep the list with your signed copy.

The post How to compare two PDF files for differences explains how changes are found. A found change isn't always an important change, and a missed change isn't always absent: the checks above cover the gaps.

Your documents stay on your device

Contracts, rent agreements and offer letters hold private details. Compare PDF has no upload step and never sends your files anywhere: your own browser reads, compares and draws them. The page contacts only this site and Google's ad service, which shows the ads that keep the tools free, and our code never gives your files to the ad code. The privacy policy has the details.