How to Compare Two Texts Online and See Every Change (Word by Word)
Sana Bano
·September 19, 2026
·8 min read
How to compare two texts online: paste both versions into a word-level diff checker and read what changed, word by word, not line by line.
To compare two texts online, paste the original on the left, the changed version on the right, and read the result: red strikethrough is what was removed, green is what was added, and anything unmarked is identical in both. The one setting that matters is granularity. For prose, compare by word or sentence, never by line.
That last point is where most people go wrong, so this post spends real time on it.
Key Takeaways
- A diff is a comparison of two texts that shows only what changed. Deletions in red, additions in green, shared text left plain.
- Line-level diffs are built for code. Re-wrap one paragraph and a line diff reports the whole paragraph as rewritten, even when zero words changed.
- Word-level diffs are built for writing. In our test, a hard-wrapped paragraph compared against the same paragraph on one line scored 100 percent similar with 0 words changed.
- Moved paragraphs should show as moved. A code diff shows a relocated paragraph as a block of deletions plus a block of additions. A prose diff labels it once.
- You can do this without uploading anything. The GPTOne diff checker runs in your browser, takes up to 200,000 characters a side, and needs no account.
What "compare two texts online" actually means
A diff checker takes two pieces of text and finds the shortest set of edits that turns one into the other. That output is the diff. The idea goes back to the Unix diff program of the 1970s, and the Wikipedia entry on diff is a good short history if you want the algorithm side.
The important thing is that every diff tool has to decide what a "unit" is before it can compare anything. A unit can be a character, a word, a sentence or a line. The tool then asks, for each unit, whether it exists in both texts.
Pick the wrong unit and the comparison is technically correct and practically useless. That is the whole problem with using a programmer's tool on an essay.
How to compare two texts online, step by step
Here is the routine we use for anything from a 200-word email to a 6,000-word article.
1. Get both versions as plain text. If they live in Word or Google Docs, open each one, select all, copy. A text compare tool works on pasted text, not on uploaded .docx or PDF files. Microsoft has its own compare and merge feature inside Word if you need formatting changes tracked too, but for the words themselves, plain text is faster.
2. Paste the older version on the left. Original, draft, before. The direction matters because the colours are relative: green means "present in the right text, absent in the left".
3. Paste the newer version on the right. Edited, final, after.
4. Set the granularity to word. Sentence mode is the other good option for prose. Line and character modes exist for code and for catching single-letter typos.
5. Read the result top to bottom. Unchanged paragraphs stay quiet. Changed paragraphs light up only on the words that moved. If a paragraph was cut and pasted elsewhere, look for the moved label rather than a deletion and a matching addition.
6. Check the stats block. Words added, words removed, sentences rewritten, paragraphs moved and a similarity percentage. That block tells you in five numbers how heavy the edit was.
The whole thing takes under a minute once you have both texts on the clipboard.
Why a line diff breaks on prose
This deserves its own section because it is the reason a purpose-built prose tool exists at all.
Code is written in lines. Each line is a meaningful unit, so a line-based comparison is the natural fit, and tools like Diffchecker were built around it. Their own FAQ says word-level precision "reads best for prose", which is an honest admission that the default view is for something else.
Prose is written in paragraphs. Where a paragraph wraps depends on the width of the box it was pasted into, the font, and whether the source added hard line breaks. None of that is a change to the writing.
We tested this. We took a 28-word paragraph, hard-wrapped one copy at 60 characters so it ran over three lines, and left the other copy as a single line. A line-based diff reports three lines deleted and one line added, which reads as a complete rewrite. The GPTOne diff checker, working at the paragraph level first and then by word inside it, reported 0 words changed and 100 percent similarity.
Same words, same order. One tool says everything changed. The other says nothing did. Only one of them is telling you something about the writing.
The same failure shows up with moved text. Drag paragraph two below paragraph three and a line diff shows a block of red where it was and a block of green where it went. We ran that too: three short paragraphs, one relocated. The prose diff produced one moved label, 0 words changed, and 100 percent similarity, which is the right answer, because the writer changed the structure and not a single word.
Reading the highlights without misreading them
A few habits make a diff much easier to trust.
Read pairs, not colours. A red word immediately followed by a green word is a substitution. "Commence" struck through, "start" in green. Treat them as one edit, not two.
Watch for punctuation-only changes. In word mode a comma added mid-sentence highlights the word it attaches to. That is not a rewrite. If a section looks heavily edited, switch to character mode for a moment and see whether it is really just commas and quotation marks.
Trust the similarity number for scale, not for judgement. The percentage is the share of words the two texts have in common, computed as a Dice coefficient. 95 percent means light copyediting. 60 percent means substantial rewriting. It does not tell you whether the rewrite was good, and it is not a plagiarism score.
Know which mode you are in. Sentence mode counts a whole sentence as changed if any word in it changed, so its similarity number runs lower than word mode on the same pair of texts. When you quote a similarity figure to someone, say which mode produced it.
Three situations where a word-level diff pays for itself
An editor sent your piece back. Rather than reading 2,000 words looking for what moved, paste your submission and their version and read only the highlights. Ten minutes becomes two, and you see the pattern in their edits, which is how you get better.
An AI tool rewrote your draft. ChatGPT, Claude and Grammarly all return a full replacement, not a redline. Paste your original and the AI version into the diff to see exactly what the model touched. If you then want to know how either version reads to an AI detector, GPTOne scans up to 50,000 characters at a time with a free account, with free credits on signup and no card required. Our post on whether Grammarly triggers AI detection goes through what tends to get flagged and why.
You need to show your revision trail. Students accused of using AI often have drafts. Diffing draft one against draft two against the final produces a record of how the essay evolved, which we cover in how to prove you didn't use AI on an essay. For the classroom-specific version, see how to compare two essays for differences.
Privacy: does the text leave your browser?
It depends entirely on the tool, and the honest answer for most free online diff tools is yes, your text goes to their server to be compared.
The GPTOne diff checker does the comparison in your browser with JavaScript. Nothing is posted to a server, nothing is stored, and there is no account, so there is no profile to attach the text to. Share links work by compressing both texts into the part of the URL after the # symbol, which browsers never send in a request. That has a consequence you should know about: the link contains the texts. Anyone you send it to can read both versions, so treat sharing the link like sending the documents.
Because there is no server step, there is no server-imposed limit either. Each side accepts up to 200,000 characters, which is roughly 30,000 words. Past that, your browser's memory is the only ceiling. For a whole book, compare chapter by chapter.
FAQ
How do I compare two texts online for free?
Paste the original in the left box and the revised version in the right box of a browser-based diff checker, set it to word or sentence mode, and read the red and green highlights. The GPTOne tool is free and works without an account.
What is the difference between a text compare tool and a diff checker?
Nothing, in practice. "Diff checker" comes from programming, "text compare" is the plain-English name. The thing to check is whether the tool compares by line (built for code) or by word and sentence (built for writing).
Can I compare two Word documents online?
Yes, by pasting their text. Open each document, select all, copy, and paste into the two boxes. If you need to compare formatting, fonts or tracked comments, use Word's built-in Compare feature instead.
Why does the diff show my whole paragraph as changed when I only fixed one word?
You are almost certainly in line mode, and the paragraph wraps differently in the two versions. Switch to word mode and the highlight shrinks to the one word you changed.
Is a similarity percentage the same as a plagiarism score?
No. Similarity between two texts you supplied tells you how much of the wording they share. Plagiarism detection compares one text against a large corpus of other people's writing. A diff checker cannot do the second job, and no honest one claims to.
Compare your two versions at gptone.me/diff-checker: free, no account, and nothing uploaded. If you want to know how a version reads to a detector afterwards, the GPTOne AI detector is one click away and comes with free credits when you sign up.