Compare two PDF versions without missing changed diagrams

Start by checking which PDF is the older revision. Then compare the wording, match pages by their headings, and inspect drawings separately. A text comparison can find a changed fee while missing a changed diagram. It can also make an unchanged section look rewritten when that section has moved.

You can follow the example with the old PDF and new PDF. Each has three pages. The old document orders its sections as Service terms, Schedule, then Appendix A. The new document orders them as Service terms, Appendix A, then Schedule.

Confirm the old and new files

Use the revision record that came with your documents to decide which file is old. A filename containing “final” or a later file modification time is not enough to establish revision order. Check the document headings before comparing pages.

That direction matters in the practice PDFs. The fee changes from USD 1,200 in the old Service terms to USD 1,500 in the new ones: an increase of USD 300, or 25% of the old fee (300 ÷ 1,200 × 100). Reverse the files and the comparison describes the change in the opposite direction.

Compare wording within matching sections

The practice Service terms contain two text changes: the fee and a change from may to must in the cancellation wording. The rest of that cancellation sentence is unchanged. See the rendered terms comparison for the old page on the left and the new page on the right, or open the section text diff to focus on the extracted wording.

A diff identifies the words that changed; it does not decide what the new wording means or whether it is acceptable. For a decision that depends on the change, record the exact old and new wording alongside the section and page numbers.

The Schedule illustrates a different problem. It moves from page 2 to page 3, but “Delivery window: October 12-16” and “Contact: Operations desk” remain unchanged. A comparison of all extracted text in page order shows the Schedule as removed and reinserted. Matching the two Schedule sections by heading shows that its position changed while those words did not. The whole-document text diff shows why a moved section needs that second look.

For this practice pair, the findings are:

Section	Old → new	What changed
Service terms	Page 1 → page 1	Fee: USD 1,200 → USD 1,500; cancellation word: may → must
Schedule	Page 2 → page 3	Section moved; the recorded delivery and contact text is unchanged
Appendix A	Page 3 → page 2	Section moved; its drawing changed while its caption stayed the same

The Schedule and Appendix A have swapped positions. Treat that as a page-order change, then assess each section’s contents on its own. Comparing old page 2 with new page 2 would put the Schedule beside Appendix A and produce a distracting list of differences. For a page-range comparison of the Schedule, pair old page 2 with new page 3.

Use Acrobat’s comparison report

In Acrobat Pro, open All tools > Compare files, assign the old and new PDFs, and use Change File or the swap control if their positions are wrong. The settings let you choose page ranges separately for the old and new files. Choose the document type that fits the layout: Reports, Spreadsheets, Magazine Layouts for continuously flowing body text; Presentation Decks to match similar pages, including moved ones; or Scanned Documents for image-based material such as drawings and illustrations. Compare text only leaves out graphic differences. Acrobat creates a report you can navigate with Go to First Change, Previous, and Next. Annotations, formatting, and background changes are off by default; select them in Filter when they matter. Save the report if you need to retain it. These are Adobe’s documented controls, not results from an Acrobat test of the practice files.

Use the report to locate changes, then check the corresponding old and new sections. If a paragraph appears in a different place, its heading and wording help distinguish a move from a rewrite.

Inspect drawings even when the text matches

Appendix A has the same caption in both practice PDFs, but its vector rectangle changes width from 100 to 160 PDF points. The rendered diagram comparison shows old Appendix A page 3 on the left and new Appendix A page 2 on the right. Those point measurements describe how the practice drawing was made. The diagram has no numeric axis, so the wider rectangle alone does not establish an increase in any real quantity or capacity.

The separate graphic-only old PDF and graphic-only new PDF make the limit of text comparison clear: their extracted strings are identical and their text diff is empty, yet their bars are visibly different. When drawings matter, inspect the corresponding pages under the same heading at a comparable zoom and orientation. Do not treat an empty text diff as evidence that the images match.

Run the local text comparison

The downloadable comparison script reads two local PDFs and prints a unified diff of their extracted nonblank lines in page order. It requires Python and PyMuPDF to be installed. From the folder containing the script and practice PDFs, run:

python compare_text.py old-version.pdf new-version.pdf > local-diff.txt

That command creates local-diff.txt on your computer; it is not a supplied download. The script makes no network call or PDF upload. You can also inspect the supplied old extracted text and new extracted text. The script performs no OCR, vector comparison, or interpretation of a document’s meaning.

Extracted text order can itself cause confusing diffs, especially in dense columns or unusual layouts. PyMuPDF’s text extraction guide explains that PDF text can come out in an unexpected order and describes Page.get_text(sort=True) for position-based sorting. The script uses position-based sorting within each page. A text diff still has no measurement of the vector rectangle.

If an expected change is absent, confirm the old and new files, the page ranges, and whether text can be extracted from the relevant pages. Check whether a comparison is set to text only or has filtered out a change type you need. For a scan with no extractable text, a blank text result does not establish that nothing changed. The practice comparison also does not examine forms, annotations, digital signatures, or attachments.

If you want to check your own review, open the known changes. The review notes CSV has blank decision fields if you need to document what a change means for your own work.

Source note: The practice PDFs were created and locally text-compared and rendered on Windows with Python 3.12.14 and PyMuPDF 1.28.2. Acrobat’s controls come from Adobe documentation reviewed October 7, 2026; Acrobat was not run in this trial.