Put two versions of the same text side by side and it shows you exactly what changed — every word added, every word removed, and every block you simply moved. No account, no upload, no waiting.
Two people edit a paragraph and swear they changed almost nothing. A contract comes back from the other side’s lawyer with “a few small tweaks”. A translation agency returns a file and you cannot see where the corrections are. Reading both versions and trying to spot the difference by eye is slow, and it is the kind of slow that misses the one number that mattered.
A text comparison tool does that reading for you. Paste the original on one side and the changed version on the other, and the difference is coloured in: struck through in red where something was taken out, highlighted in green where something was put in. This one compares by word, by character or by whole line, tells you how similar the two texts are as a percentage, and — unusually for a free diff checker — notices when a paragraph was moved rather than rewritten, so a reordered document does not read as if it was rewritten from scratch.
Three steps, and the first result appears before you have finished pasting.
Paste it, or drag a file straight onto the box. Plain text formats all work — .txt, .md, .csv, .tsv, .json, .xml, .html, subtitle files and source code in most languages. Nothing is uploaded; the file is read by your own browser.
The comparison runs as you type, so there is no button to find. Got them the wrong way round? Swap flips the two sides without you re-pasting anything.
Every change is coloured and counted, with a similarity score at the top. Step through them one at a time with Alt + ↓, then copy the report or download it as a text file to send to whoever asked what changed.
The same two texts, compared three ways. Most online diff tools give you one.
The similarity percentage follows whichever you pick, because the honest answer depends on the question. ABC-1042 against ABC-1043 scores 50% by word — the code matches, the number does not — and 87.5% by character, because seven of its eight characters are unchanged. Both are true. Which one you want depends on whether you are checking a reference or proofreading a digit.
Three switches that turn noise back into signal.
The switches only change what counts as a match. Your text is never altered, never re-wrapped and never cleaned up — what you pasted is what you copy back out.
Almost every free difference checker gets this wrong, and it is the change people most often need to see.
Move a clause from the middle of a contract to the end and an ordinary diff reports it twice: a deletion where it used to be, an insertion where it now is, and a similarity score that drops as though you rewrote it. Nothing was rewritten. It moved.
This tool checks whether a removed block reappears, unchanged, somewhere else in the other version. When it does, both places are marked in amber as a moved block and counted once. A reordered document then reads the way it actually is — the same words in a different order — instead of looking like a rewrite.
It is the difference between a report saying “47 changes” and a report saying “you reordered two paragraphs and changed one date”.
What the tool marks when a contract comes back “with a few small tweaks”.
The Supplier shall deliver the Goods within 30 days of the Order Date. Payment falls due 30 days after delivery. Either party may terminate on 60 days’ written notice.
The Supplier shall use reasonable endeavours to deliver the Goods within 45 days of the Order Date. Payment falls due 30 days after delivery. Either party may terminate on 30 days’ written notice.
Three changes, and the two that matter most are single numbers. Word view marks 30 → 45, 60 → 30 and the inserted hedge use reasonable endeavours to — the phrase that quietly turns a hard obligation into an effort. This is the reading a comparison tool is for: the alterations that survive a skim.
The tool does not care what the text is. These are the jobs it gets used for.
Find what the other side changed before you sign, and check that a revised policy says what the summary email claims it says.
Compare a paragraph before and after an edit — including one you ran through our paraphrasing tool or humanizer — and see precisely what moved.
A file comes back “lightly edited”. Line view tells you in a second whether that is true.
Two configuration files, two API responses, two exports of the same table. Line view keeps the structure; character view finds the one changed value.
Check a reviewer’s corrections to a translated file, or two versions of an .srt, without scrolling through both.
See how far two model answers diverge, or what a second pass through an AI writer actually altered.
Half the people searching for this want a download. Here is the honest comparison, including where this tool is the wrong choice.
| Option | Best when | Costs you | Watch out for |
|---|---|---|---|
| This page | You want an answer now, on any device, without installing anything — and the text is prose, a policy, a config file or a data row | Nothing. No account, no limit | Needs the page to load once. Not built for repository-sized files |
| Notepad++ with the Compare plugin | You are on Windows, work with files on disk all day, and need it to run with no internet at all | An install, and Windows only | Line-level only — it will not show you which word inside a line changed |
| VS Code (Compare Active Editor With…) | The files are already open in a project you are working in | A code editor you may not otherwise want | Heavy for two paragraphs of an email. Aimed at code, not prose |
The diff command | You are scripting it, comparing hundreds of files, or working over SSH | Comfort with a terminal | Output is for machines first. No word-level view without extra flags |
| Word’s Compare Documents | Both versions really are .docx and you need tracked changes back out | Microsoft Word | Compares formatting as well as text, which is often noise |
If your text is already in a browser tab — and most text is — the online route is simply the shortest path. If you are diffing a source tree, use the tools built for it.
A number that is easy to quote and easy to misuse.
The score is the share of tokens the two texts have in common: twice the matching tokens divided by the combined length of both sides. At word level, two texts sharing 90 of their 100 words score 90%. At character level it counts characters instead, which is why a single changed digit barely moves it. Switch the comparison level and the number changes, because you have changed the question.
What it is not: a plagiarism score. It compares the two texts you pasted and nothing else — it has no index of the web, no database of student work, and no opinion about who wrote either version. Two people given the same brief will produce texts with a high word overlap without either copying the other. If you need to think about that properly, our academic integrity guide covers where writing help ends and misconduct starts, and the AI detector answers a different question again — whether text reads as machine-written — with its own limits stated on that page.
Treated as what it is — a measure of overlap between two specific texts — it is genuinely useful: it turns “quite a lot changed” into a number you can put in an email.
This matters more here than on most tools: people compare contracts, medical letters, unreleased copy and internal policies. Our privacy policy sets out how the rest of the site handles text, and which tools do send text to a model — this one does not.
The things people ask before they trust a comparison.
Paste the original into the first box and the changed version into the second. The comparison runs as you type — there is no button to hunt for. Removed words are struck through in red on the left, added words are highlighted in green on the right, and the line numbers down each side carry a − or + so the changes are still readable if you cannot distinguish the colours. You can also drop two files straight onto the boxes.
Yes, and there is no account, no sign-in and no daily limit. The comparison is arithmetic rather than AI, so it costs us nothing to run and there is nothing to meter. Every plan on our pricing page, including the guest one, gets the whole tool.
No. The comparison engine is a JavaScript file your browser already downloaded with the page, so it runs on your own machine. That is the same privacy you would get from a desktop app, without installing one.
Once the page has loaded, yes — the comparison itself never contacts a server, so it keeps working if your connection drops. Loading the page in the first place needs the internet, so if you want a tool that works from a cold start on a plane, a desktop application is still the right answer.
Notepad++ with the Compare plugin does the same job well and works entirely offline, but it is Windows-only and needs installing. This runs anywhere with a browser, including a phone or a locked-down work laptop, and adds word-level and character-level views that the plugin does not have.
You can — open both files, press Ctrl + Shift + P and choose File: Compare Active Editor With…. VS Code is excellent for code and unbeatable when the files are already in a project you have open. It is a heavier answer for two paragraphs of an email.
Anything plain text: .txt, .md, .csv, .tsv, .json, .xml, .html, .srt, .vtt, and source files in most languages. Files are read in your browser and capped at 4 MB each.
Not directly — a .docx or .pdf is a compressed container, not text, so opening one here would show you machine formatting rather than your writing. Copy the text out of both documents and paste it in. Our word counter reads .docx if you need the text extracted first.
Yes. On a narrow screen the two texts stack, and the tabs above the workspace switch between entering text and reading the differences.
The share of tokens the two texts have in common, at whatever level you are comparing — words, characters or whole lines. It is a description of overlap, nothing more: it is not a plagiarism score and it is not evidence about who wrote what.
No. Nothing you paste or drop here is uploaded, logged or kept. It exists in the tab until you close it. That is a property of how the tool is built, not a policy promise — the comparison happens in code running on your own device, which our privacy policy describes in full.
Yes. Line view is usually what you want for code and data, because it keeps the structure obvious; character view is better for spotting a changed digit in an ID or a stray comma in a CSV row.
Text Compare sits alongside eleven other tools. Everything free stays free.
Saved work lives in your drafts, and what each plan includes is on the pricing page. Text Compare needs neither — it is free on every plan, including without an account.
Upgrade