W
Write It

Explainer

Does Turnitin detect ChatGPT?

Turnitin reports an estimated percentage of a document that reads as machine-written. What that number is, and is not, deserves a straight answer.

Write It teamUpdated 3 min read

Short answer: Turnitin has an AI writing indicator, it produces a percentage, and teachers can see it. Longer answer: what that percentage means is more complicated than either side of this argument usually admits.

What Turnitin actually does

Turnitin has two separate things. The similarity report, which it has run for two decades, matches your text against a database of published work and previously submitted papers — that is plagiarism detection and it works by comparison.

The AI writing indicator is newer and works completely differently. It has nothing to compare against, so it analyzes the statistical properties of the writing itself and estimates what proportion appears machine-generated.

What it measures

Broadly the same signals every detector uses. How predictable each word is given the preceding text, and how much sentence length and structure vary across the document. Models pick likely words and write evenly; people are lumpier.

Which means the indicator is not detecting AI. It is detecting statistical regularity, and inferring AI from it. Those are different things, and the gap between them is where every false positive lives.

How reliable is it?

Turnitin publishes its own accuracy figures, measured under its own conditions. Independent testing consistently finds a lower real-world rate, particularly on shorter documents and on certain kinds of writing.

  • Non-native English writers are flagged more often. Simpler vocabulary and more regular sentence structure read as low perplexity. This is a documented, unresolved problem across the whole category.
  • Formal academic register is flagged more. Because it is deliberately uniform.
  • Short passages are least reliable. There is not enough text to measure variation.
  • Heavy editing can increase the score. Polishing smooths out exactly the irregularity detectors look for.

What teachers see

A percentage, and highlighted passages the model considers most likely machine-written. Whether an institution enables the AI indicator at all is up to the institution, so some teachers see it and some do not.

Turnitin’s own guidance is that the score is an indicator requiring human review, not evidence of misconduct. Institutions that treat the number as proof are using it against the vendor’s stated guidance.

If you are wrongly flagged

  1. Do not panic and do not confess to something you did not do

    A flag is a question, not a finding.

  2. Open your version history

    A Google Doc showing the paper written over days, with revisions and abandoned sentences, is the strongest evidence available to you.

  3. Gather your notes, outlines, and sources

    Anything showing process. Drafts, reading notes, a messy outline.

  4. Ask what specifically was flagged

    You are entitled to know which passages and why. If a flagged paragraph is one you can explain in detail, say so.

  5. Ask about the appeals process

    Most institutions have one. Use it.

See your draft before your teacher does

Install Write It and run the AI Detector on your own writing to find what reads as machine-written.

Add Write It to ChromeSee plans

Frequently asked questions

Can Turnitin tell if I used ChatGPT?
It can produce an estimate that a passage reads as machine-written, based on statistical properties of the text. It has no record of what tool you used and cannot know — it is inferring from the writing itself.
Does Turnitin detect paraphrased AI text?
Rewriting changes the statistical properties the indicator measures, so a rewrite can change the score. This is not something any tool can promise a result on, and treating it as a reliable technique is a bad bet.
Can I check my paper on Turnitin myself before submitting?
Usually not — Turnitin is licensed to institutions rather than individuals. Some schools give students access through a draft submission folder. Other detectors are available directly, though they will not necessarily agree with Turnitin.
Will this get past an AI detector?
No tool can promise a particular detector score, and we do not make that promise. The humanizer rewrites AI-flavored phrasing so it reads naturally to a person; the AI Detector shows you what still reads as machine-written so you can decide what to do about it.

Keep reading