Review a Contract Revision with a Text Diff

A revised contract rarely announces its important changes. The email says "minor updates throughout," the document looks the same length, and the one altered clause sits quietly on page nine. Re-reading the whole agreement to find it is slow, and skimming risks missing exactly the change you most needed to see. A text comparison — a diff — solves this by showing only what changed, so your attention goes to the differences instead of the ninety percent that stayed identical.

Text and Document Comparison compares two texts side by side or inline, at line, word or character level, and can export the result as an HTML report. This post builds a review workflow around it, using a fictional agreement, and shows how the choice of diff mode at each stage catches different kinds of change.

Stage one: the line-level sweep

Start with a line-level comparison. Paste the original agreement on one side and the revised version on the other, and run the diff with line mode. This gives you the shape of the revision: which paragraphs were added, which were removed, and which were touched at all. Most of the document will show as unchanged, and that is useful information — it tells you where not to spend your time.

For a fictional example, consider a services agreement between a designer, Ana, and a client company at example.org. The revised version is twelve pages. The line-level diff shows that ten pages are untouched, one paragraph was added to the payment section, and three paragraphs were modified in the termination section. Ana now knows her review is really about four paragraphs, not twelve pages. That narrowing is the first benefit of the workflow, and it is available before she has read a single changed word.

Resist the urge to start reading the changed paragraphs immediately. First, note the summary counts: how many lines were added, removed and changed. A revision described as "minor" that adds forty lines has a different character from one that changes four words, and the counts give you a sense of scale before the detail pulls you in.

Stage two: the word-level read

Now switch to word-level mode for the changed paragraphs, and read only those. Line mode tells you a paragraph changed; word mode tells you how. In Ana's agreement, the termination paragraph looks similar at a glance, but the word-level diff highlights three changed words: the notice period moved from "thirty" to "fourteen" days, "written notice" became "notice," and a sentence about return of materials gained the word "promptly."

Each of those is small and each matters. Fourteen days instead of thirty changes Ana's planning horizon. Dropping "written" widens what counts as notice. Adding "promptly" introduces a vague standard where a precise one might have served her better. None of these changes would leap out of a casual re-read, because the surrounding sentences are identical and the eye slides over familiar text. The diff stops that slide by highlighting exactly the words that differ.

This is the stage to slow down and take notes. For each highlighted change, write one line: what changed, and whether you accept it. That note becomes your reply to the other side, and it is far more useful than a general impression that the revision "looks fine."

Stage three: the character check on numbers and names

Finally, run a character-level comparison on the passages that contain numbers, dates, names and amounts. Word mode can hide a changed digit inside an otherwise unchanged word — a fee moving from $4,800 to $4,300, or a date moving from the 15th to the 18th, may read as a changed word without making the size of the change obvious. Character mode highlights the individual digits, so a single changed figure stands out even in a long paragraph of identical text.

In Ana's agreement, the character check catches one more change the word read had softened: the late-payment fee changed from 2 percent to 5 percent, buried in a paragraph about invoicing that the line sweep had flagged but that Ana had skimmed. Character mode is tedious across a whole document, which is why it belongs last and only on the passages that carry figures. Used selectively, it is the safety net under the whole review.

For guidance on comparing documents generally, see the guide to comparing documents.

Worked example: the four-paragraph review

Here is Ana's review, summarised as the note she actually sends back. Payment section, one added paragraph: a new clause requiring invoices within five days of month end — acceptable, since she invoices promptly anyway. Termination section, three changed words: notice period shortened from thirty to fourteen days — she asks for twenty-one as a compromise; "written notice" reduced to "notice" — she asks for "written" to be restored, so there is a record; "promptly" added to the materials-return sentence — she asks what timescale is meant.

Invoicing paragraph, one changed figure: late-payment fee raised from 2 to 5 percent — acceptable, since it only applies to late payment. Total review time: under half an hour, for a twelve-page agreement, with a reply that addresses every change specifically. Without the diff, the same review would have meant re-reading twelve pages and hoping the eye caught the four altered passages among the familiar ones.

Exporting the comparison as an HTML report gives Ana a record she can keep alongside the agreement. If the other side later disputes what changed, the report shows it, rather than relying on memory of a read-through.

Honest limitations

A diff shows textual differences, not meaning. It cannot tell you whether a changed clause is fair, enforceable, or consistent with the rest of the agreement — that judgement still needs a human, and for important contracts, a lawyer. A diff also depends on clean input: if the two versions have different formatting, line breaks or spacing, the comparison may flag cosmetic differences alongside real ones, and options to ignore whitespace and case exist precisely to reduce that noise. Finally, a diff of plain text cannot compare scanned images or heavily formatted documents directly; the text must be extractable first.

FAQs

Should I use line, word or character mode?

Use all three, in stages. Line mode maps the revision, word mode reads the changed passages, and character mode checks numbers and names. Each mode answers a different question, and together they cover the document.

Can a diff catch a single changed number?

Yes, especially in character mode, where a changed digit is highlighted individually. Word mode will also flag the changed figure as a changed word, but character mode makes the exact digit obvious.

What if the comparison flags formatting differences I do not care about?

Use the options to ignore whitespace and case differences. That strips out cosmetic noise — extra spaces, line-break changes, capitalisation — and leaves the substantive wording changes visible.

Can I share the comparison result with someone else?

Yes. The comparison can be exported as an HTML report, which shows the differences with summary counts and can be sent or filed alongside the documents.

For comparing structured data rather than prose, JSON Studio compares JSON structurally, and Compare PDF diffs two PDFs page by page.

human review pending client designation