Can Turnitin Detect ChatGPT, Claude and Gemini in 2026?

Turnitin says its AI writing detection can identify text produced by ChatGPT and many other large language models. As of November 2026, its English-language FAQ names GPT, Claude, Gemini, Llama, Grok, DeepSeek, Mistral and Amazon Nova models. The result is a separate AI percentage shown to instructors, not students, and Turnitin says it should not be the sole basis for action.

Key points

• The AI writing percentage and the similarity score are completely independent.
• An AI score needs at least 300 words of long-form prose, no more than 30,000 words and a .docx, .pdf, .txt or .rtf file (Turnitin).
• Scores from 1% to 19% are shown as an asterisk, not a number (Turnitin).
• Turnitin says it aims to keep false positives under 1% for documents with over 20% AI writing (FAQ).
• For English, Turnitin says it also flags AI text altered by paraphrasing and "bypasser" (humanizer) tools (FAQ).

Which AI models does Turnitin say it can detect?

Turnitin publishes the list of large language models its English detector covers. As of November 2026, the FAQ says the English model can detect content from these models, the newest dated March 2026:

Turnitin describes ChatGPT as a writing tool based on GPT models, so the list names model versions, not app releases. The FAQ says Turnitin is "continuously updating our model to detect these additional LLMs." Its October 14, 2026 English update improved detection "while minimizing false positives" but names no new models. Other supported languages use separate models with their own coverage.

Can Turnitin detect ChatGPT 5?

Turnitin lists nine GPT-5 series models, from GPT-5 to GPT-5.4-pro, and no GPT model dated after March 2026. Being listed does not mean every passage is caught: Turnitin says that to keep false positives low it "might miss some AI written text."

Can Turnitin detect Claude AI?

Turnitin says yes for the seven Claude versions it lists, from Claude 3 Haiku to Claude Opus 4.6. Its documentation does not say how the detector performs on Claude or Gemini versions that are not listed.

How does Turnitin detect AI writing?

Turnitin scores each sentence and reports the share that crosses a threshold. Each sentence gets a score between 0 and 1 "based on the sentence in context with the surrounding text," and the document score is the percentage of sentences above the threshold (FAQ).

Only prose counts. The model removes text it predicts is not real sentences, "such as lists, headers, references, data tables, and text that appears to be formulas or code."

Turnitin calls its system "a modern deep learning approach based on the transformer architecture," also trained on its in-house corpus of academic data. It says the model is "not explicitly programmed" to measure signals such as perplexity or burstiness, so predictions "may not always be explainable" feature by feature. In October 2026 it updated the English model to improve accuracy where text switches between human and likely AI writing. For the general technique, see how AI detectors work.

Does Turnitin detect AI or just plagiarism?

It produces both, as two separate results that "are completely independent and do not influence each other," according to the FAQ. AI writing highlights are not visible in the Similarity Report.

Turnitin's guide for students is precise: "Turnitin does not check for plagiarism." It compares a submission with its database and highlights matches, including correctly quoted and referenced text. Students see a Similarity Report only if their instructor allows it.

For what each AI percentage band means, see our guide to Turnitin AI scores.

What does a paper need to get an AI score?

Turnitin's file requirements for an AI Writing Report are:

• File size less than 100 MB
• At least 300 words of prose text in a long-form writing format
• No more than 30,000 words
• English, Spanish, Japanese, Arabic, German, Dutch or Turkish
• A .docx, .pdf, .txt or .rtf file

Files that miss these show a gray indicator with no percentage. Turnitin says the model does not reliably detect AI text in non-prose "such as poetry, scripts, or code," and does not detect short-form writing such as bullet points, tables or annotated bibliographies (Turnitin). In documents of only a few hundred words, the FAQ warns, some mixed text "could be flagged as entirely AI-generated."

Why do low Turnitin AI scores show an asterisk?

Turnitin hides scores between 1% and 19% to reduce false positives. Since a July 2024 release, "no score or highlights are attributed" in that range and the indicator shows an asterisk. Turnitin says results in this range "are less reliable and false positives are more likely." Reports generated before the change may still show a number below 20%.

What is Turnitin's AI detection false positive rate?

Turnitin says it aims to keep the false positive rate "under 1% for documents with over 20% of AI writing," or about one in every 100 fully human-written documents (FAQ). Before every update it tests on over 700,000 academic papers written before ChatGPT was released.

Three conditions apply:

• It covers documents with over 20% AI writing. Below that, the score is hidden.
• It is a document-level rate. In a June 2023 post, Turnitin put the sentence-level rate at "around 4%," with 54% of those sentences "right next to actual AI writing." The current FAQ gives no updated sentence-level figure.
• It trades off missed AI text. A document scored at 50% "could contain as much as 65% AI writing."

The FAQ lists text that can trigger false positives: content "without a lot of structural variation, text that literally repeats itself, or text that has been paraphrased without developing new ideas."

On non-native writers, Turnitin's own October 2023 study found the gap between English language learners and native writers "small, and not statistically significant" for documents of 300 words or more. Each group's rate was slightly above the 1% target but not significantly different from it. For shorter documents, the gap was larger.

What independent research says

Independent studies predate Turnitin's current model and mostly test other detectors, so they show general limits rather than Turnitin's accuracy today.

• Weber-Wulff and colleagues tested 12 publicly available tools plus two commercial systems, Turnitin and PlagiarismCheck. They concluded that the available tools were "neither accurate nor reliable," with a bias toward labelling output as human-written, and that "content obfuscation techniques significantly worsen the performance of tools" (arXiv, June 2023).
• Liang and colleagues (Patterns, July 2023) ran seven widely used GPT detectors on 91 TOEFL essays and 88 US eighth-grade essays. The US essays were classified accurately, but the TOEFL essays had an "average false-positive rate: 61.3%." Turnitin says its detector was not included in that evaluation.
• Institutions differ. In August 2023, Vanderbilt University said it would disable Turnitin's AI detector "for the foreseeable future," citing false positives and too little information on how it works.

Can Turnitin detect ChatGPT if you paraphrase it?

Turnitin says yes for English when the paraphrasing is done by AI tools, including humanizers, which it calls bypassers. Its timeline:

1. December 6, 2023: it began detecting likely AI text "even if it may have been paraphrased using an AI word spinner" (release notes).
2. July 16, 2024: the report added a category for AI-generated text that was AI-paraphrased, naming Quillbot as an example tool.
3. August 27, 2025: it added bypasser detection for text that "may have been intentionally modified by AI humanizer tools," trained and tested for English only.
4. August 4, 2026: the report merged the categories into one set of blue highlights, as the signals separating them "have become more difficult to interpret."

Turnitin says it is "unable to disclose the names of these tools," and the documents cited here give no separate accuracy figures for this detection. See also whether Turnitin detects humanized AI.

What can students see of the AI report?

Students do not see the AI score. The FAQ says the indicator and report "are not visible to students," though instructors can download the report as a PDF and share it. An existing score changes "only if it's re-submitted."

Not every school has the feature: the FAQ says it is only available to customers that license Turnitin Originality (or an iThenticate 2.0 add-on), and administrators can turn it off. For what you can check yourself, see whether you can check your paper on Turnitin before submitting.

What does Turnitin say the score should be used for?

A starting point, not a verdict. The FAQ says the percentage "should not be used as the sole basis for action or a definitive grading measure by instructors," and Turnitin "does not make a determination of misconduct."

Its report guide says the model "may not always be accurate" and that misconduct decisions need "further scrutiny and human judgment" under the institution's policies. Its review guidance calls the score "a single data point rather than a definitive response." If you are flagged for your own work, see what to do if you are falsely accused of using AI.

Disclosure: this article is published by Ryne AI, which makes an AI Detector. Ryne's AI Report runs Turnitin's AI detector on your paper for you.

Frequently asked questions

Can Turnitin detect ChatGPT copy and paste?

The standard AI Writing Report scores the text in the submitted file; its documentation does not describe recording how text was typed or pasted. Paste tracking is part of Turnitin Clarity assignments, where students write inside Turnitin's writing space and a flag appears when pasted words make up more than 50% of the final word count. Clarity work done outside that space is not tracked and cannot be submitted for a Writing Report.

Does Turnitin detect Grammarly as AI?

Not for ordinary grammar fixes, says the FAQ: in its tests on human-written documents, grammar-checker changes were not flagged "in most cases." Grammarly's generative features, such as drafting and paraphrasing, are different: "Content produced using these features will likely be flagged as AI-generated by our detector."

Can Turnitin detect AI in a PowerPoint?

Turnitin's AI file requirements list only .docx, .pdf, .txt and .rtf, so PowerPoint is not an accepted type. Bullet-point text also falls outside the long-form prose the model reliably scores.

Can Turnitin detect AI on a PDF?

Yes. PDF is an accepted type if the file is under 100 MB with at least 300 words of prose and no more than 30,000 words. Turnitin recommends that instructors have all students submit in the same format.

Sources

1. Turnitin: Turnitin's AI writing detection capabilities FAQs (accessed November 2026)
2. Turnitin: File requirements for an AI Writing Report (accessed November 2026)
3. Turnitin: Using the AI Writing Report (accessed November 2026)
4. Turnitin: AI writing detection model (accessed November 2026)
5. Turnitin: Turnitin release notes (accessed November 2026)
6. Turnitin: Turnitin product updates (accessed November 2026)
7. Turnitin: How to access the AI Writing Report (accessed November 2026)
8. Turnitin: How should I review the AI Writing report? (accessed November 2026)
9. Turnitin: Understanding the similarity score for students (accessed November 2026)
10. Turnitin: Reviewing the Writing Report (accessed November 2026)
11. Turnitin: FAQs for Instructors using Turnitin Clarity (accessed November 2026)
12. Turnitin: Understanding the false positive rate for sentences of our AI writing detection capability (14 June 2023)
13. Turnitin: New research: Turnitin's AI detector shows no statistically significant bias against English Language Learners (26 October 2023)
14. Turnitin: Turnitin Expands Capabilities Amid Rising Threats Posed by AI Bypassers (27 August 2025)
15. arXiv: Weber-Wulff et al., Testing of Detection Tools for AI-Generated Text (21 June 2023, revised 10 July 2023)
16. Patterns (PMC): Liang et al., GPT detectors are biased against non-native English writers (10 July 2023)
17. Vanderbilt University: Guidance on AI Detection and Why We're Disabling Turnitin's AI Detector (16 August 2023)

Turnitin is a trademark of Turnitin, LLC. Ryne is an independent service and is not affiliated with or endorsed by Turnitin.