What the Compare PDFs tool does
Compare PDFs takes two PDFs, pulls the text out of both, and shows you line by line what changed — what got added, what got removed, what's the same. Drop the original in one slot and the updated version in the other, click Compare, and you get a redline view in a few seconds. Pick Side-by-side for the classic two-pane layout that contract reviewers expect, or pick Unified for a single column with +/− markers (the format git uses for code diffs).
One thing to say out loud, because this is the question most people would ask first: this is a TEXT diff, not a visual diff. The tool reads the text content stream from each PDF and compares those character strings. It does not compare layout, fonts, colors, images, signatures, or where things sit on the page. If you need pixel-level visual comparison, Adobe Acrobat's Compare Files feature is the right tool and we'll say that again later in the article. For comparing the words in two PDFs — which is what 90% of contract revision, proofreading, and policy-monitoring work needs — this tool is the right shape and runs in seconds.
The whole pipeline runs in your browser. PDF.js (Mozilla's renderer, the same one that ships inside Firefox) extracts the text. The diff package — same algorithm used by git and most code-review tools — runs the line-level comparison. The PDFs never upload.
How the diff works, in plain language
The algorithm is called diffLines and the idea is simple: find the longest common subsequence of lines between the two documents, mark everything else as either "added" (lines that appear in the updated PDF but not the original) or "removed" (lines that appear in the original but not the updated), and emit the result in source order. It's the same algorithm computer scientists have been refining since the 1970s and it's what GitHub uses every time you open a pull request.
What counts as a "change"? Any line that isn't byte-for-byte identical. Change a single word in a paragraph, and the entire line gets marked removed AND the new version gets marked added — that's standard diff behavior. Whitespace counts: a trailing space changes the line. Punctuation counts: swapping a comma for a semicolon changes the line. If you need word-level granularity — only the changed words highlighted inside an otherwise unchanged sentence — open both files in a dedicated redline tool like Litera Compare or Workshare Compare. That's a different algorithm and a much heavier product.
For most real revisions — the counterparty edited the indemnification clause, the editor rewrote the second paragraph, the policy team swapped "shall" for "will" in three places — line-level is enough to see exactly what moved. The summary at the top tells you the raw numbers: X added, Y removed, Z unchanged. The body shows you which lines.
Side-by-side vs. unified — which view to use
The tool offers both views because different jobs want different shapes.
| View | What you see | Best for |
|---|---|---|
| Side-by-side | Two columns. Left = original. Right = updated. Removed and added chunks are paired into the same row when they sit adjacent in the diff stream. | Contract revisions, prose where you want to read the old version and the new version in parallel and feel the texture of the change. |
| Unified | One column. Removed lines prefixed with − (red). Added lines prefixed with + (green). Context lines unchanged. Same format git diff emits. | Quick scanning, sharing screenshots with developers, anyone fluent in git review. |
Both views feed off the same diff data, so toggling doesn't re-run anything. Flip back and forth as needed — most reviewers settle on side-by-side for the first pass (to feel the shape of the changes) and unified for the second pass (to count exact additions and removals).
How to compare two PDFs
- Drop your two PDFs into the slots. The "Original" slot is the earlier version. The "Updated" slot is the newer one. Order matters: removed lines are ones in Original that aren't in Updated; added lines are the reverse. Swap the slots and "added" becomes "removed."
- Click Compare. The tool reads each PDF (a few seconds per hundred pages), extracts the text with PDF.js, reconstructs line breaks from the layout, and runs
diffLines. - Read the summary at the top: X added, Y removed, Z unchanged. That number alone often tells you what you need — "only 4 lines changed across 80 pages" is a different conversation than "412 lines changed."
- Scroll the diff. Toggle between Side-by-side and Unified depending on what your eyes prefer.
- Need to share what you found? Take a screenshot — there's no download because this is a viewer, not an editor. Most reviewers screenshot a few key chunks and paste them into Slack, email, or the contract management system.
A worked example: a contract that came back redlined
You sent a SaaS subscription agreement to a counterparty named Globex on April 3. On May 8 they returned it with their proposed revisions. The cover email says "minor changes" but you've been around long enough to want to verify that yourself.
You drop globex-msa-2026-04-03.pdf into Original and globex-msa-2026-05-08-revised.pdf into Updated. Click Compare. The summary reads: 23 added, 18 removed, 1,847 unchanged. So about 41 lines moved across what looks like a 60-page document. "Minor" is doing some work in that cover email.
You scan the side-by-side view. The first batch of changes is housekeeping — addresses got updated, an effective date got bumped. Section 4 (Fees) has three edits where Globex changed the payment terms from Net 30 to Net 45 — that's a real change worth flagging. Section 7 (Limitation of Liability) has eight lines removed and twelve added; the cap got rewritten and the carve-outs got expanded. That's the one that needs a careful read. Section 12 (Termination) has two minor word swaps that don't change meaning.
You screenshot the Section 7 diff, paste it into the Slack channel with your general counsel, write "Globex rewrote 7. Worth a 10-minute look." Total time from receiving the redline to surfacing the substantive change: about four minutes. The alternative — reading both PDFs side by side in two windows of your screen, mentally diffing 60 pages of legal prose — is the half-day version of the same job.
Why text-only and not visual
The first question people ask when they hear "PDF comparison" is "does it find moved blocks and font changes?" Honest answer: no, and that's deliberate.
A real visual diff is a serious piece of software. Rendering both PDFs to pixels, aligning corresponding pages, detecting blocks that moved, color-coding font and color changes, distinguishing meaningful edits from layout reflows — that's months of engineering work and the field is dominated by paid products (Adobe Acrobat Compare Files, Litera Compare, Workshare Compare, Draftable). They charge for it because the work is hard and the customers are law firms with serious budgets.
What we observed is that most people who reach for "PDF compare" don't need pixel-level analysis. They have two versions of a contract, a policy, a research paper, or a memo, and they want to know what the words say differently. The text content of a PDF is straightforward to extract, the diff algorithm is well-understood, and the result is faster to skim than a multi-color visual redline. Build the 90% tool that opens in a browser, run it free, and point people at Acrobat for the 10% of cases that genuinely need visual comparison. That's the trade.
What the tool catches and what it misses
Worth being precise about, because the value of any diff tool is knowing what it can tell you.
| Change type | Caught? | How it shows up |
|---|---|---|
| Word added or removed in a sentence | Yes | Whole line marked removed + new line marked added |
| Paragraph inserted or deleted | Yes | Block of added or removed lines |
| Numbers, dates, dollar amounts changed | Yes | Line-level swap, same as text |
| Whitespace difference (trailing space, double space) | Yes | Line marked changed |
| Font changed (Times → Arial) | No | Same text → not flagged |
| Color changed (black → red) | No | Same text → not flagged |
| Image inserted, removed, or moved | No | Images aren't in the text stream |
| Signature added on page 12 | No | Signatures are images |
| Block moved from page 3 to page 7 | Partially | Shows up as a removal on page 3 and an addition on page 7 |
| Page reordering with same content | Partially | Same as block moves |
| Layout change (margins, columns) with same text | No | Same text → not flagged |
The pattern: if a change affects the words, this tool finds it. If a change only affects how the words look, this tool doesn't. For contract review, prose proofreading, and policy monitoring that's almost always the right scope. For brand-guideline audits, layout review, and signature verification it isn't.
Common use cases
Real jobs people show up with.
- Contract revisions. A counterparty returns a contract with their edits. You want to confirm what they actually changed, beyond whatever they highlighted in the cover email. Diff in 30 seconds; review only the substantive changes.
- Proofreading drafts. Comparing draft N to draft N+1 to verify your edits landed and nothing else moved. Especially useful when multiple people are editing the same document.
- Policy and terms monitoring. Was that privacy policy updated quietly? Was the company's code of conduct rewritten? Save the PDF once a month, diff against last month's copy. You'll see exactly what changed.
- Translation review. Same document in two regional variants — US English vs UK English, Brazilian vs European Portuguese. The diff surfaces every translated phrase that differs.
- Year-over-year financial statements. Same template, different fiscal year. The diff shows you which line items grew, which shrunk, which appeared, which disappeared.
- Academic paper revisions. A reviewer returns a paper with edits; you want a fast count of how heavily they revised it before reading it line by line.
- Compliance attestations. Many compliance docs are version-controlled. Diff the current published version against the last attested version to confirm what changed since the last audit.
About Adobe Acrobat Compare Files
Adobe Acrobat Pro has a Compare Files feature that does what this tool does plus visual comparison. It catches font changes, image edits, layout shifts, color changes, the lot. It costs $20-30 a month depending on the plan and requires Acrobat Pro installed locally. If you're doing high-volume legal redline work, brand-asset comparisons, or anything where visual fidelity matters as much as text content, Acrobat is the right answer and we'd recommend it without hesitation.
What this tool offers instead: drag two files, see the text diff in seconds, free, no install, no upload. For the prose-comparison case — which is most cases — that trade is good. The smart workflow is to use the free text diff first to triage the change scope, then pull out Acrobat if and only if the visual layer matters.
The privacy story
The PDFs you compare are often the documents you most don't want to share. Contracts. Sealed court filings. Drafts of M&A paperwork. Compensation reports. Personal medical records. Internal investigation memos. These are the documents where "we upload to a server and process it there" — the design that iLovePDF, SmallPDF, and most cloud-based PDF compare tools use — is the wrong shape.
Compare PDFs runs entirely in your browser. PDF.js extracts the text in JavaScript memory. The diff package runs the comparison in JavaScript memory. The results render to the DOM. The PDFs never touch a network. There's no server to honor a privacy policy because there's no server in the path. Open your browser's developer tools, go to the Network tab, run a full comparison: zero outbound requests during compare. The page load is the only network activity.
That's not a privacy promise — it's an architecture. There's nothing to leak because there's nothing in flight.
Related PDF tools
- Extract Text from PDF — pull the raw text out of a PDF, useful when you want to diff in a dedicated text editor or feed the content into another tool.
- Text Diff Checker — the same line-level diff for plain text input rather than PDFs. Paste two text blocks, see the comparison.
- Split PDF — break large PDFs into chapters before diffing; section-by-section is faster than running a multi-thousand-page comparison in one pass.
- PDF Merger — combine the diff result (as a screenshot) with the original PDFs into a single review packet.
- Sign PDF — when you've reviewed the revised contract and you're ready to sign it.
About Microapp
Microapp is a collection of small, focused tools that open in a tab, do one job, and let you leave. Compare PDFs is one of about 140 tools in the catalog. The membership model is Costco-shaped — members get clean pages and AI at compute cost; non-members get the same tools with ads on the page. 10% of every dollar Microapp earns goes to charity, off the top, audited quarterly. The tools are picked, not just aggregated.
Frequently asked questions
What counts as a "change"?
Any line that isn't byte-for-byte identical. The tool splits both PDFs into lines, runs the standard diffLines algorithm, and marks each line as added, removed, or unchanged. Changing a single word marks the whole line removed AND adds the new version — that's how line-level diffs work. Whitespace counts: a trailing space changes the line. For word-level granularity, use a dedicated redline tool like Litera Compare.
Why text-only? Doesn't a PDF have layout?
Because honest is faster than half-magic. A real visual diff is a serious product — rendering both PDFs to pixels, aligning pages, detecting moved blocks, color-coding font changes. Adobe Acrobat's Compare Files feature does that, and it costs money for a reason. We compare the extracted text, which catches what 90% of contract-revision and proofreading work needs. Layout, fonts, images, signatures — none of those are detected. We say so plainly above every result.
How does the line-level diff work?
We use the diffLines algorithm from the diff package — the same approach git and most code-review tools use. It finds the longest common subsequence of lines and marks the rest as added or removed. In Side-by-side view, adjacent removed and added chunks are paired into the same row. In Unified view, removed lines come before their corresponding added lines, prefixed with − and +.
Will it work on large PDFs?
Up to 100 MB and 500 pages per file. Text extraction is the slow step (a few seconds per hundred pages on a normal laptop); the diff itself is near-instant for documents under ~50,000 lines. For multi-thousand-page legal-discovery work, split each PDF into chapters first with our Split PDF tool and diff section by section.
Does this work on scanned PDFs?
No. Scanned PDFs are images of text, not text. The extractor returns empty or near-empty results, and the diff would be meaningless. If your PDFs are scans, run them through an OCR tool first (Adobe Acrobat, macOS Preview, Tesseract) to add a text layer, then compare the OCR'd versions here.
Are my PDFs really not uploaded?
Correct. PDF.js (the library that renders PDFs inside Firefox) and the diff algorithm both run in your browser. Your files go from disk into browser memory, get diffed, and the result renders on screen — never to a server. Check the network tab while comparing: zero outbound requests.
Why use this instead of Adobe Acrobat Compare Files?
Acrobat's Compare Files is the gold standard for visual PDF comparison — page rendering, font detection, image diffing — and if you need that, use it. This tool is for the text-only case: faster (drag two files, see results in seconds), free, no software install, runs in any browser. For pure prose comparison — contracts, articles, policy docs — text-only is usually what you want and is faster to skim than a visual redline.