Product
Semantic Redlining: How Pactum’s Track Changes Engine Works
June 2026 · 5 min read · By Evans Selasi Adika
Standard diff tools break on formatted legal text. Pactum’s semantic redline engine compares contracts at the word level — green for additions, red strikethrough for deletions. Accept or reject each change individually.
The Problem with Traditional Track Changes
When a lawyer asks an AI to revise a clause, the AI often rewrites the entire paragraph. The output looks clean — but what exactly changed? Which words were added? Which were deleted? Was a critical limitation of liability quietly removed?
Standard text diff tools (like those in Word) compare formatting as well as content. A change in font size or paragraph spacing is treated the same as a change in legal substance. The result: noise that obscures the signal.
How Pactum’s Redline Engine Works
Step 1: Snapshot. Click “Snapshot” to capture the current state of the contract. This creates a baseline.
Step 2: Edit. Make your changes — manually or by applying AI suggestions. Edit as much as you want.
Step 3: Compare. Click “Compare Against Snapshot.” Pactum’s engine performs a word-level diff with a 20-word lookahead.
Step 4: Review. Added words appear in green. Deleted words appear in red with strikethrough. Each change is listed in the Redline panel with Accept and Reject buttons.
Why Word-Level Matters
The difference between “within 30 days” and “within 14 days” is two words. In a traditional diff, the entire sentence might be flagged as “modified.” In Pactum’s semantic redline, only “30” (deleted, red) and “14” (added, green) are highlighted. The lawyer sees exactly what changed and nothing else.
Key Detail
The redline engine runs entirely in your browser.
No contract text is sent to any server for comparison. The snapshot and the current version are compared locally in JavaScript.