Anyone who has pasted two paragraphs into a "compare text" tool and gotten back a wall of red and green with no clear signal has run into the same underlying problem: naive text comparison. A tool that simply compares character-by-character, or line-by-line with exact matching only, tends to mark far more as "changed" than actually changed — because the moment one word shifts position, everything after it looks different too.
What LCS actually means for a diff
The standard, correct approach to text diffing is based on the longest common subsequence (LCS) problem: finding the longest sequence of lines (or words) that appear, in order, in both texts. Everything in that common subsequence is marked unchanged; everything else is either an addition (present in the new text, not the common subsequence) or a removal (present in the old text, not the common subsequence). This is the same conceptual foundation behind tools like GNU diff and the diff view in Git.
The practical benefit is that a diff computed this way finds the smallest, most meaningful set of changes rather than the most obvious one. If you insert a new sentence in the middle of a paragraph, a good LCS diff shows just that sentence as added — the surrounding text stays marked unchanged, because it still matches.
Line-level vs. word-level
Line-level diffing is the right starting point for most documents — it tells you which lines were added, removed, or touched. But when a single line is edited (say, one word changed in a sentence), showing the whole line as "removed" and its replacement as "added" isn't very informative. A more useful diff pairs up a removed line with its most likely replacement and runs a second, word-level LCS diff between them, so you can see precisely which words were added or removed within that line — not just that "something changed here."
Common false signals to watch for
- Trailing whitespace — a line that differs only by trailing spaces will show as fully changed unless the tool has a whitespace-insensitive mode
- Case changes — "Total" vs "total" registers as a difference in a case-sensitive diff, which may or may not be what you want to flag
- Reordered content — LCS diffs assume content stays roughly in order; heavily reordered documents can still produce a busy-looking diff even with a correct algorithm
Where this is useful
Comparing two drafts of a contract or policy document, checking what changed between two versions of a config file, verifying an AI-generated rewrite only touched what you asked it to touch, or reviewing edits a collaborator made to shared notes. Since the comparison runs entirely in your browser, it's also a reasonable option for text you'd rather not paste into a third-party web service.
Reading a diff efficiently
- Scan for red/green pairs first — these are usually edits to existing lines, not wholesale additions or deletions
- Solid green blocks with no adjacent red are pure additions; solid red blocks with no adjacent green are pure removals
- Turn on "ignore whitespace" if you're comparing code or exported text where indentation may differ but content doesn't
A diff tool is only as trustworthy as the algorithm behind it. An LCS-based line diff with word-level detail on changed lines gives you a comparison you can actually rely on for real editorial or technical review.
Frequently Asked Questions
It uses a real LCS (longest common subsequence) algorithm at the line level, so only lines that actually changed are highlighted — unchanged lines stay marked as unchanged.
Yes. When word-level highlighting is enabled, modified lines are further diffed word-by-word so you can see exactly which words were added or removed, not just that the line differs.
Yes — the Text Diff Checker highlights additions in green and removals struck through in red, with line numbers on both sides. It's a one-time $4.99 purchase — no subscription, no account required.
Yes, there are toggles to ignore leading/trailing whitespace per line and to ignore case, useful when formatting differences shouldn't count as real changes.
No. The entire diff is computed locally in your browser — neither text block is sent to any server.