The short answer
Yes. Turnitin has an AI writing detection feature that estimates how much of a submission was likely generated by an AI tool such as ChatGPT. It appears as an AI writing indicator next to the Similarity Report. Three conditions decide whether it runs on your paper:
- Your institution has to license it and switch it on. Turnitin says it only processes submissions for AI writing when the institution has the feature enabled, and administrators can turn it off.
- The file has to qualify: at least 300 words of prose, in a supported language and file type.
- The result is an estimate. Turnitin itself says the percentage “should not be used as the sole basis for adverse actions against a student”.
What Turnitin’s AI detection checks
According to Turnitin’s AI writing detection FAQ, a submission is split into overlapping segments of sentences. A model gives each segment a value between 0 and 1 for how likely it is to be AI-generated, each sentence takes the score of the segments it sits in, and the sentence scores are combined into one percentage for the document.
Two details in that description matter more than they look:
- It only counts “qualifying text”. Turnitin defines this as prose sentences in long-form writing, such as an essay or a dissertation. It says the model does not reliably detect AI text in poetry, scripts or code, or in bullet points, tables and annotated bibliographies. So the percentage is a share of the prose, not of the whole file.
- It is separate from the similarity score. Turnitin states that the two numbers are “completely independent and do not influence each other”. Similarity is text matching against a database. The AI percentage is a prediction about how the text was produced.
Turnitin lists the models its English detector covers, including recent GPT, Gemini, Claude and Llama versions, and says the list will keep growing. Our guide on how AI detectors work explains the general method behind this kind of classifier.
Word count, languages and file types
Turnitin’s file requirements for an AI Writing Report are short and specific. If a submission misses any of them, no AI percentage is produced.
| Requirement | What Turnitin states |
|---|---|
| Minimum length | At least 300 words of prose text in a long-form writing format |
| Maximum length | No more than 30,000 words |
| Languages | English, Spanish, Japanese and Arabic |
| File types | .docx, .pdf, .txt, .rtf |
| File size | Less than 100 MB |
If a paper is written in a language that is not supported, Turnitin says the detector will not process it and no report is generated. That does not mean the paper is unchecked: the Similarity Report still runs, and your teacher still reads the work.
How the score is shown
Instructors see an indicator that can be in one of several states. These come from Turnitin’s guide to using the AI Writing Report.
| What the indicator shows | What it means, per Turnitin |
|---|---|
| 0% | No qualifying text was identified as likely AI-generated. |
| *% (an asterisk) | Something between 1% and 19% was detected. No number and no highlights are shown, because results in this range are less reliable and false positives are more likely. |
| 20% to 100% | The share of qualifying text identified as likely AI-generated, including likely AI text that was then changed with a paraphrasing tool. The passages are highlighted in the report. |
| Gray, no percentage | The submission could not be processed, for example because it does not meet the file requirements or was submitted before the feature was enabled. |
| Error icon | Processing failed. Turnitin suggests trying again later. |
A worked example
Suppose a 2,000-word report has 1,500 words of paragraphs and 500 words of tables and bullet lists, and the indicator shows 40%. That is 40% of the 1,500 words of prose, about 600 words, not 40% of the file. It also does not mean “40% certain”. It means the model predicts that roughly that share of the prose was likely generated by AI, and the instructor is expected to read the highlighted passages and decide what they show.
What Turnitin says about false positives
A false positive is human writing flagged as AI. Turnitin’s FAQ puts a number on it: the company says it keeps the false positive rate under 1% for documents with more than 20% AI writing, and explains that as possibly flagging one fully human-written document in every 100. It says it tests each model update on more than 700,000 academic papers written before ChatGPT existed.
The same page is open about the limits:
- Low scores are unreliable. That is the stated reason for hiding everything from 1% to 19% behind an asterisk.
- Some human writing is more likely to be flagged. Turnitin names text without much structural variation, text that repeats itself, and text that has been paraphrased without developing new ideas.
- Short documents are “all or nothing”. With only a few hundred words, Turnitin says the prediction is made on a single segment, so a mix of AI and original text could be flagged as entirely AI-generated.
- It misses things too. To keep false positives low, Turnitin accepts that it may not catch all AI text. Its example: a document reported as 50% could contain as much as 65%.
Keep in mind that the 1% figure is Turnitin’s own measurement of its own product. We have not tested it and cannot confirm it. What is not in dispute is the conclusion Turnitin draws: the detector “may not always be accurate”, and deciding whether misconduct happened takes human judgment and the institution’s policy.
Paraphrasing tools and “humanizers”
A common belief is that running AI text through a paraphrasing tool makes it safe. Turnitin’s documentation says otherwise. Its English detector also looks for likely AI-generated text that was then modified by an AI paraphrasing tool, a word spinner or a “bypasser” (often sold as a humanizer), and it runs automatically on every submission where AI detection is enabled. Turnitin does not publish which tools it can detect.
One limit is stated plainly: this paraphrasing and bypasser detection is currently available for English submissions only. Turnitin says its Spanish and Japanese detectors do not include it yet.
Whether a tool hides AI text from a detector is the wrong question anyway. If a course does not allow AI-written work, rewording it does not make it yours, and disguising it is usually treated as worse than the original shortcut. Rewording also changes nothing about what your teacher can see without software: drafts, version history, and whether you can explain your own argument. If AI is allowed in your course, the honest route is to disclose it and cite it. See is using AI plagiarism? and how to cite ChatGPT.
Who sees the result
Turnitin states that the AI writing indicator and report are not visible to students. Instructors and administrators see them, and an instructor can download the report as a PDF and share it. That is why you may see a similarity score on your submission and nothing about AI.
Other points from the same FAQ that vary by institution or setup:
- The feature is available to institutions that license Turnitin Originality, or as an add-on in iThenticate 2.0. Not every Turnitin customer has it.
- Inside an LMS such as Canvas, Moodle or Blackboard, the AI score is not shown in the LMS screens. The instructor opens the Turnitin report to see it. Our guide on AI detection in Canvas covers that setup.
- Work submitted before the feature was enabled has no AI score unless it is submitted again.
- There is no separate repository for AI detection, according to Turnitin.
If your work is flagged
Turnitin’s advice to instructors on what to do when the AI percentage is high is to start from the institution’s policy and use the score to open a conversation, not as a punishment. The final decision, it says, rests with the instructor. If you are the student in that conversation:
- Ask to see the report. You cannot see it in your own account, so ask for the PDF and look at which passages are highlighted.
- Bring your process. Version history in Google Docs or Word, outlines, notes and earlier drafts show how the text developed.
- Be ready to talk about the work. Explaining your argument and your sources is strong evidence that you wrote it.
- Say what tools you used. If you used a grammar checker or a translation tool, say so, and check what your course policy allows.
- Know the procedure. Your institution’s academic integrity policy sets out the steps and how to appeal.
Teachers will find a fuller process in how teachers check for AI.
Checking your own draft first
Students usually cannot run Turnitin’s AI detection on their own. It is sold to institutions, and the result is only shown to instructors. You can still check a draft with an independent tool, as long as you are clear about what that tells you.
The Plagiarism Checker Plus AI detector gives a sentence-level view of which parts of a text read as AI-like. The first check is free without an account, up to 1,000 words. A free account includes 3,000 words a month. The text needs at least 50 words, and it is never used to train AI models.
It is a different detector from Turnitin’s, trained differently, so the two can disagree. No tool can tell you in advance what Turnitin will show, and you should be wary of any service that claims it can. Use a check like this to find passages worth rereading, and keep your drafts, which protect you better than any score.