The short answer
Yes. Turnitin detects AI writing through an AI writing indicator inside its Similarity Report, but only when your institution licenses Turnitin Originality and switches the feature on. Instructors and administrators see the score; students do not. It needs 300 to 30,000 words of prose in English, Spanish, Japanese or Arabic, and hides scores below 20%.
Does Turnitin check for AI on every paper?
No. Turnitin checks for AI only where the institution has paid for it and left it switched on. According to Turnitin's AI writing FAQ, AI writing detection is available only to customers that license Turnitin Originality. Schools on Turnitin Similarity or Feedback Studio alone have to speak to their account manager to get it.
Administrators can also turn it off. The same FAQ says admins can enable or disable the AI writing feature from their settings page, and that Turnitin will not process a submission for AI writing if the institution does not use the feature. So two students submitting to Turnitin at different universities may get very different treatment: one paper is scored for AI, the other is not.
How the detector works, in Turnitin's words
Turnitin splits a submission into overlapping segments of sentences. Each segment gets a score between 0 and 1 for how likely it is to be AI-generated, each sentence inherits the scores of the segments it sits in, and those sentence scores are combined into one document percentage. The FAQ describes the current model as a transformer-based deep learning system trained on Turnitin's own corpus of academic writing.
Turnitin says the model is not programmed to measure single signals such as "perplexity" or "burstiness". It learns patterns from training data instead, which also means, in Turnitin's words, that individual predictions "may not always be explainable in simple feature-by-feature terms". As of October 2026 its FAQ lists GPT-5.x, Gemini 3, Claude Sonnet 4.5 and 4.6, Llama 4, DeepSeek and Grok models among those it was built to detect, plus tools based on them.
Can students see their Turnitin AI score?
Not by default. Turnitin's FAQ states that only instructors and administrators can see the AI writing indicator, and that the indicator and report "are not visible to students". An instructor can download the AI report as a PDF and share it with you, but that is their choice.
The AI score is also kept apart from the similarity score. The two are calculated independently and do not influence each other, and Turnitin says there is no AI score shown directly in the learning management system (LMS) interface. Instructors have to open the Similarity Report and select the AI Writing tab to see it.
If your instructor raises an AI concern, you are entitled to ask to see the report they are relying on. Our guide on what to do if you are falsely accused of using AI covers how to ask for it and what evidence of your writing process helps.
What do the Turnitin AI percentage and asterisk mean?
The percentage is the share of qualifying text, meaning prose sentences in long-form writing, that the model thinks was likely generated by AI or generated and then altered by a paraphraser or bypasser tool. It is not a percentage of the whole file. If your document has tables, lists or code, those parts are left out of the calculation, so the highlights and the number may not appear to line up.
Since a July 2024 update, scores between 1% and 19% are not shown. Turnitin's report guide says its testing "found that there is a higher incidence of false positives when the percentage is between 0 and 19", so the indicator shows an asterisk (*%) and no highlights instead. Reports generated before 8 July 2024 may still show a number below 20%.
| What you see | What it means |
|---|---|
| 0% | No qualifying text was identified as likely AI-generated or AI-altered. |
| *% (asterisk) | AI was detected above 0% but below 20%. No percentage or highlights are shown because false positives are more likely in this range. |
| 20% to 100% | That share of the qualifying prose was identified as likely AI-generated, or AI-generated and then paraphrased or bypassed. Matching text is highlighted. |
| Gray, no percentage (- -) | The file could not be processed for AI: it predates the feature or does not meet the file requirements. |
| Error (!) | Processing failed. Turnitin suggests resubmitting and contacting support if it persists. |
What can Turnitin's AI detector not check?
A paper only gets an AI score if it meets Turnitin's file requirements for an AI Writing Report. Anything outside these limits produces the gray "no percentage" state rather than a score.
| Requirement | Turnitin's limit |
|---|---|
| Minimum length | 300 words of prose in a long-form writing format |
| Maximum length | 30,000 words |
| File size | Less than 100 MB |
| File types | .docx, .pdf, .txt, .rtf |
| Languages | English, Spanish, Japanese, Arabic |
| Paraphraser and bypasser detection | English submissions only |
Turnitin is also open about the kinds of writing it does not score reliably. Its guide says the model does not reliably detect AI in poetry, scripts or code, and does not detect it in short-form or unconventional writing such as bullet points, tables or annotated bibliographies. Since August 2023, bibliographies are excluded from AI processing altogether.
Short papers carry an extra caveat. The FAQ explains that with only a few hundred words the prediction is mostly "all or nothing", because the model is scoring a single segment. A short piece that mixes your writing with AI text could therefore be flagged as entirely AI-generated.
Can Turnitin detect paraphrased or humanized AI text?
Turnitin says it can, in English. The AI writing model release notes record three steps. In December 2023 it began detecting AI text run through word spinners. In July 2024 the report gained a separate category for AI-generated text that was then AI-paraphrased, with QuillBot given as an example. On 27 August 2025 it added detection of AI bypasser tools, also called humanizers, which rewrite AI output to look more human.
Turnitin will not say which paraphrasers or bypassers it was trained on, on the grounds that a list would make evasion easier. Its Spanish and Japanese detectors do not include paraphraser or bypasser detection at all. Turnitin has kept updating the model, with releases in October 2025 and February 2026 aimed at improving recall, and a Spanish-model update in May 2026.
What about Grammarly?
Turnitin's FAQ says its tests on human-written documents found that, in most cases, spelling, grammar and punctuation changes made by Grammarly and similar tools were not flagged as AI. Text produced by Grammarly's generative features, such as drafting, paraphrasing and summarizing, is a different matter: Turnitin says that content will likely be flagged.
How accurate is Turnitin's AI detection?
Turnitin's stated target is a document-level false positive rate under 1% for documents with over 20% AI writing, meaning roughly one wrongly flagged paper in every 100 fully human-written ones. Before each model release it says it tests on more than 700,000 academic papers written before ChatGPT existed. It also admits a trade-off: to keep false positives low it may miss AI text, so a document scored at 50% AI "could contain as much as 65% AI writing".
The sentence level is less certain. In a May 2023 update, Turnitin put its sentence-level false positive rate at about 4%, and reported that 54% of falsely flagged sentences sat right next to genuinely AI-written text. In other words, a highlighted sentence in a mixed document is not proof that that particular sentence was machine-written.
What independent testing shows
In Plagino's March 2026 benchmark of 1,000 documents (500 AI-written, 500 human-written, default settings, document-level scoring), Turnitin caught 98.2% of the AI documents and wrongly flagged 1.4% of the human ones. That detection rate was the highest of the seven tools we tested, ahead of Plagino's own 97.4%; Plagino's false positive rate was lower, at 0.8%. One corpus is not every classroom, so read the confidence intervals on the benchmark page before leaning on either number.
Detectors in general have a documented weak spot with second-language writers. A 2023 Stanford-led study found that several AI detectors misclassified more than half of TOEFL essays by non-native English speakers as AI-generated. Turnitin says it included second-language writers in its training data to reduce this risk, but every detector, including Turnitin's, can be wrong. Our explainer on why AI detectors produce false positives goes into the causes.
Which universities have turned off Turnitin's AI detector?
Several institutions kept Turnitin for plagiarism checking but disabled the AI indicator. Vanderbilt University switched it off in August 2023, estimating that even a 1% false positive rate on its roughly 75,000 annual submissions could mean about 750 papers wrongly labeled. It also cited bias against non-native English writers and a lack of transparency about how detection works.
The University of Waterloo's Turnitin page states that AI detection "will no longer be available to Waterloo users as of September 2025". Curtin University announced in September 2025 that it would disable Turnitin's AI writing detection from 1 January 2026, while keeping text-matching originality checks active.
Turnitin itself says the AI percentage "should not be used as the sole basis for adverse actions against a student", and that it does not make a determination of misconduct. Whether or not your school uses the indicator, the decision rests with people applying the institution's own academic integrity policy.
What should students do before and after submitting?
You cannot run Turnitin's AI check on your own work, because there is no individual Turnitin account. If you want to see which of your sentences an automated detector is likely to flag, you can run your draft through a separate tool such as the Plagino AI detector. It will not reproduce your school's Turnitin score, and our guide to the best alternative to Turnitin compares the options, including what each one checks against.
The more useful habit is keeping evidence of your process. Save outlines, notes and drafts, and write in a tool with version history. If an AI flag is ever raised, that record shows how the work developed, and it matters far more than any second detector's score. Ask your instructor which AI uses your course allows before you start, not after.
Frequently asked questions
Does Turnitin detect ChatGPT?
Yes, where the institution has AI writing detection enabled. Turnitin's FAQ lists GPT-4o, GPT-5 and later GPT-5.x models among those its English detector was built to detect, along with models from Google, Anthropic, Meta and others, and tools built on them.
Can students see the Turnitin AI score?
No, not by default. Turnitin says the AI writing indicator and report are visible only to instructors and administrators. An instructor can download the AI report as a PDF and choose to share it with a student.
What does an asterisk mean on a Turnitin AI report?
It means Turnitin detected some AI writing above 0% but below 20%. Turnitin hides the exact figure and highlights in that range because its testing found false positives are more common at low percentages.
What is the minimum word count for Turnitin AI detection?
A submission needs at least 300 words of prose in a long-form writing format, and no more than 30,000 words. Files must be under 100 MB and in .docx, .pdf, .txt or .rtf format.
Does Turnitin flag Grammarly as AI?
Usually not for spelling, grammar and punctuation fixes. Turnitin says its tests found those edits were not flagged in most cases. Text written with Grammarly's generative features, such as drafting or paraphrasing, will likely be flagged.
Is a Turnitin AI score proof of cheating?
No. Turnitin states that its AI percentage should not be the sole basis for adverse action against a student and that it does not decide whether misconduct happened. That decision belongs to the institution under its own policy.