How to Compare Two Essays for Differences (Drafts, Edits, Rewrites)

Compare two essays for differences in under a minute: two drafts, two submissions, or an original vs an AI rewrite. Word-level highlights, nothing uploaded.

To compare two essays for differences, paste the earlier version on the left and the later one on the right of a word-level diff tool, and read the highlights: red for cut, green for added, plain for shared. There are three cases people mean by this: two drafts of the same essay, two different students' submissions, and an original against a ChatGPT or Grammarly rewrite. The same tool handles all three, but you read the result differently each time.

One thing this post is not: a guide to writing a compare-and-contrast essay. If that is what you searched for, the UNC Writing Center's guide is the best one online. This post is about mechanically comparing two finished texts.

Key Takeaways

• Three use cases, one tool. Draft vs final, submission vs submission, original vs AI rewrite. The text compare tool for essays at GPTOne has a mode for each.
• We ran a real example. A 260-word draft against its 223-word final: 61 words added, 98 removed, 13 sentences rewritten, 67.1 percent similarity in word mode.
• Similarity is shared words over total words. It is a measure of how much wording two texts have in common, not a plagiarism verdict and not an AI verdict.
• Nothing is uploaded. The comparison runs in the browser, so a student's essay never leaves the teacher's machine. There is no account and no server-side limit.
• The diff shows what changed, not who wrote it. For the AI question, either version goes to an AI detector, which needs a free account.

Case 1: comparing two drafts of your own essay

This is the most common reason anyone compares two essays, and the least discussed. You have a first draft and a revised one. You want to know what you actually changed, whether the argument survived, and whether the word count dropped for the right reasons.

Here is a real run. We wrote a 260-word first draft of a short opinion essay on teenagers and social media, then revised it the way a decent editor would: cut the throat-clearing, tightened the sentences, swapped two body paragraphs, and deleted the summary paragraph at the end. Then we pasted both into the GPTOne diff checker in word mode.

The stats block came back like this:

Read that as an editor would. A third of the wording changed. Far more was removed than added, which is what tightening looks like. Thirteen sentences rewritten across six paragraphs means the revision touched almost everything, and the one paragraph that came through untouched was the one we had left alone on purpose.

Now the part that matters for essay writers specifically. In sentence mode, the same two texts scored 26.5 percent similarity. That is not a contradiction. Sentence mode counts a whole sentence as changed when any word in it changed, and this revision touched a word or two in most sentences. Word mode tells you how much wording survived. Sentence mode tells you how many sentences survived intact. For a draft-to-final comparison, word mode is the honest number, and it is the one to quote.

Case 2: comparing two students' essays

Teachers compare two submissions for one of two reasons: two essays look suspiciously alike, or one student's draft and final look suspiciously unalike.

For the first, paste one submission left and the other right. Shared passages appear unmarked. If two essays written independently share three consecutive sentences, that is worth a conversation. If they share a topic sentence and a quotation from the assigned reading, that is what an assignment produces.

For the second, paste the draft left and the final right, using the Student drafts mode. What you are looking for is the shape of the change. A student who revised their own work leaves a trail that looks like our example above: sentences tightened, a paragraph moved, transitions rewritten, the argument intact. A final that shares almost nothing with the draft, with a similarity in the teens, is a different document, and the question becomes where it came from.

Two things to hold onto here.

First, the diff is evidence to discuss, not a verdict to deliver. Students rewrite from scratch for legitimate reasons, and a similarity number cannot tell you why. Google's own version history in Docs is a complementary record if the student wrote there, and Grammarly now offers an Authorship report that tracks typing and pasting. Neither replaces talking to the student.

Second, privacy. Most online comparison tools, and every plagiarism checker, send the essays to a server. With GPTOne's diff checker the comparison happens in your browser, nothing is stored, and there is no login, so a student's work never leaves your device. If you are working under FERPA or a similar policy, that is the difference between a tool you can use and one you cannot.

Our post on the best AI detector for teachers in 2026 covers the detection side of the same job.

Case 3: comparing an essay with its AI rewrite

ChatGPT, Claude, Gemini and Grammarly's generative features all do the same thing when you ask them to improve an essay: they return a full replacement. None of them shows you a redline. So most people never find out how much the tool actually changed.

Paste your original left and the AI version right, switch to the Original vs AI rewrite mode, and you find out. The stats block gives you words swapped, sentences rewritten and a similarity figure. In our experience light "fix my grammar" requests land above 90 percent similar. "Make this sound more academic" often lands in the 50s, which means the model rewrote half your essay while you thought it was polishing.

That figure has practical consequences. Institutions increasingly draw the line between assistive editing and generative rewriting, and we went through where Turnitin puts that line in does using Grammarly count as AI on Turnitin. The diff is how you find out which side of it your essay is on.

Once you know what changed, the second question is how each version reads to a detector. The GPTOne AI detector scores text from ChatGPT, Claude, Gemini, DeepSeek and other models at 99.99 percent accuracy on text, up to 50,000 characters per scan. It needs a free account, which comes with free credits and no card. Run both versions. If the rewrite scores high and the original does not, you have your answer about which sentences to take back.

How to read an essay diff without fooling yourself

Set granularity to word before you look at anything. In line mode, a paragraph that wraps differently between two pastes shows as entirely rewritten. Word mode compares paragraph to paragraph first and then highlights only the words that moved. Our companion post on how to compare two texts online shows the test where the same paragraph scored 100 percent in word mode and looked fully rewritten in line mode.

Look for the moved label. When a paragraph is cut and pasted somewhere else without edits, the diff marks it as moved rather than as a deletion plus an addition, and the similarity stays at 100 percent. If the paragraph was moved and edited, it shows as changed in its new position, which is also correct.

Read substitutions as pairs. Red word, green word, next to each other. That is one edit.

Quote the mode with the number. "67 percent similar in word mode" is information. "67 percent similar" on its own invites the wrong argument, because the other person may have sentence mode in their head.

Keep the essay as plain text. The tool compares pasted text. Open the Word or Docs file, select all, copy, paste. Formatting is not compared, which for essays is a feature.

FAQ

How do I compare two essays for similarity?

Paste one essay in each box of a browser-based diff checker set to word mode. Unmarked text is shared between the two. The similarity percentage summarises how much of the wording they have in common.

Can a diff tool tell me if a student used AI?

No. A diff shows what changed between two texts you supplied. To estimate whether a text was AI-generated you need an AI detector, and even then the score is a probability to discuss with the student, not proof.

What is a good similarity percentage between a draft and a final essay?

There is no correct number. Light copyediting typically stays above 85 percent. A serious structural revision can drop to 60 or below while still being the student's own work. The pattern of edits tells you more than the figure.

Does the essay get uploaded when I compare it?

Not with GPTOne's diff checker. The comparison runs in your browser and nothing is sent to a server or stored. Other tools vary, and many free ones do process the text server-side, so check before pasting student work.

Can I compare two essays saved as Word documents?

Yes, by pasting the text. Open each document, select all, copy, paste into the two boxes. Word's own Compare feature is the right tool if you also need to see formatting and comment changes.

Paste your two essays into the GPTOne diff checker and read the stats block. When the question turns from "what changed" to "who wrote it", the AI detector is one click away with free credits on signup.