Text Diff Checker

What does a diff checker actually show you?

A diff checker shows what changed between two versions of a text, and the useful ones show it precisely. Knowing that "line 12 is different" is almost worthless in a long paragraph; knowing that thirty became sixty is the answer you came for.

That is the difference between comparing by line and comparing inside the line. A line-level tool paints the whole line red and the whole line green, leaving you to spot the edit by eye. This one pairs the edited line with its original and compares again inside it, so the highlight lands on the exact words, or the exact characters, that moved.

Take an invoice number. INV-2026-0431 against INV-2026-0413 is one transposed digit, and a line-level diff cannot point at it. Here, only the 31 and the 13 light up.

Word, character, or auto?

Three levels of precision, because different text wants different granularity:

• Word marks whole words as changed. Best for prose, contracts and documentation, where a half-highlighted word is harder to read than a whole one.
• Character marks only the characters that differ. This is what catches a transposed digit, a changed file extension, or a single letter in an identifier.
• Auto starts at word level, then drops to characters whenever two swapped words clearly share most of their content. It is the right default for mixed text.

Switching level never recompares from scratch in a way you notice, so it costs nothing to try both on the same pair.

Side by side or unified?

Side by side puts the two versions in parallel columns, which is how people read documents: old on the left, new on the right, aligned line for line. Unified stacks the old line directly above the new one, the way code review tools do, which is easier to follow on a narrow screen and keeps related changes together.

Neither is better in general. Prose and contracts read well side by side; code and config read well unified. Both are available here at no cost, which is worth saying plainly because several comparison tools charge for one of them.

How do I read the statistics?

Four numbers sit above the result, and the one worth understanding is modified.

An edited line is one line that changed, so it is counted once as modified. Lines that exist on only one side are counted as added or removed. Many diff tools have no modified category at all and record an edit as one removal plus one addition, which doubles the total and makes a two-word fix look like four changes.

The similarity percentage is the share of characters the two texts have in common overall. It answers a question you usually have before reading anything: is this a light edit or a rewrite?

How does the comparison work?

In three passes. First, the two texts are split into lines and compared with a longest-common-subsequence algorithm, the same idea behind the Unix diff command, which finds the largest set of lines that appear in both, in order.

Second, each run of removed lines is paired with the run of added lines that follows it. When two paired lines are similar enough, they become a single modified line instead of a separate deletion and insertion.

Third, every modified line is compared again internally, by word or by character, and the result is passed through a readability cleanup. That last step matters more than it sounds: a raw character comparison of "mouse" against "sofas" produces a jumble of single letters, and the cleanup collapses it into one clean swap of the whole word. Everything happens in your browser, and a long line is detected and handled at word level so the page never freezes.

What this tool does not do

It compares text you paste, and nothing else. It will not open PDF, Word or PowerPoint files, it will not save a comparison behind a shareable link, and it does not colour code by syntax.

Those are deliberate omissions, not things we forgot. Parsing office formats needs a heavy library that would slow the page for everyone, and saving a diff would mean storing your text on a server, which this tool never does. If your text lives in a document, open it, copy the part you care about, and paste it here; the comparison works exactly the same.

When is it worth using?

Any time two versions of something exist and the difference matters. Checking which clause a counterparty quietly edited in a returned contract. Confirming a config file changed in only the one line you meant. Comparing two answers from an AI to see what a prompt tweak actually did. Proofreading an edited draft against the original before it goes out.

Because the comparison runs locally and nothing is uploaded, logged or stored, none of that material leaves your device, which matters when the text is a client contract or an internal config rather than a blog post.