Home / Guides

Can Turnitin detect ChatGPT, Claude, and Grammarly?

Turnitin publishes a surprising amount about its AI detector if you know where to look. This guide sticks to what Turnitin's own documentation says, plus what independent sources add, so you know what the tool claims and where it's weak.

Updated October 10, 2026 · By the IsThisAI? team

What Turnitin's AI score is

Turnitin's AI writing indicator sits inside the Similarity Report and shows the percentage of "qualifying text" that its model thinks was likely generated by AI, or generated by AI and then changed by a paraphraser or bypasser tool (Turnitin FAQ). A few things about that number surprise people:

• It's separate from the Similarity score. The two don't influence each other. A paper can have 0% similarity and a high AI score, or the reverse.
• Only long-form prose counts. Lists, bullet points, code, and other non-sentence text are ignored, so the percentage isn't necessarily a share of the whole document.
• Scores under 20% aren't shown. From 1% to 19% you'll see only an asterisk, because false positives are more likely there.
• Students usually don't see it. Only instructors and admins see the indicator, though instructors can download the report as a PDF.

Turnitin also says the score "should not be used as the sole basis for action" and that it "does not make a determination of misconduct" (Turnitin FAQ).

Under the hood, Turnitin splits a submission into overlapping segments, scores each one between 0 and 1, and combines those into sentence scores and an overall percentage. It says its model is a transformer-based deep learning system trained on academic writing, and that it isn't explicitly programmed to measure things like "burstiness" or "perplexity" (Turnitin FAQ).

ChatGPT vs. Claude (and Gemini, Llama, Grok)

Turnitin keeps a list of models its English detector "can detect content from." As of its October 2026 FAQ, the list includes OpenAI models from GPT-4o through GPT-5.4, Google's Gemini models through Gemini 3.1 Pro preview, Anthropic's Claude models including Claude Sonnet 4.5 and 4.6, Claude Haiku 4.5, and Claude Opus 4.5, plus Llama, Mistral, DeepSeek, Amazon Nova, and xAI's Grok 4.1, "and tools based on these LLMs as well" (Turnitin FAQ).

So the honest answer to "can Turnitin detect Claude?" is: Turnitin says yes, for the Claude versions on its list. Two caveats matter:

• Turnitin doesn't publish a separate catch rate for each model. Its headline number is about false positives (under 1% at the document level for documents with over 20% AI writing), not about how much AI text it catches. It openly says it may miss some AI text to keep false positives low.
• Newer models arrive constantly. Turnitin says it keeps expanding coverage and that results may change as the detector is updated (Turnitin FAQ).

Independent research has repeatedly found that detectors struggle once AI text is edited or paraphrased (Weber-Wulff et al., 2023), which is one reason a detector catching "ChatGPT" in a lab doesn't mean it catches every lightly edited essay in practice.

Grammarly edits

Turnitin answers this one directly. Its detector "is not tuned to target Grammarly-generated spelling, grammar, and punctuation modifications," and in its tests on human-written documents, "in most cases" changes made by Grammarly (free and premium) or other grammar checkers "were not flagged as AI-written." It then adds that this excludes Grammarly's generative AI features, such as "draft generation, paraphrasing, summarizing, and other features," and that content made with those "will likely be flagged as AI-generated" (Turnitin FAQ).

In short: fixing commas is fine; asking a tool to rewrite or write your paragraphs is the risky part. Our full guide, Will Grammarly get flagged as AI?, goes deeper.

PDFs, PowerPoint, and short assignments

For an English submission to get an AI score, Turnitin lists these requirements (Turnitin FAQ, Turnitin Guides):

• File types: .docx, .pdf, .txt, or .rtf
• At least 300 words of prose in a long-form format, and no more than 30,000 words
• Under 100 MB
• A supported language

PDFs are fully supported. PowerPoint files aren't on that list, and slide decks are mostly short bullet points anyway, which Turnitin doesn't analyze. If a slide deck is converted to PDF, only sections that read as real paragraphs would count toward the 300 words. Short discussion posts under 300 words won't get an AI score at all.

AI humanizers and paraphrasers

"Humanizer" tools rewrite AI text to dodge detectors. Turnitin says its English detector now also looks for "AI-generated text that may have been modified by AI paraphraser or bypasser (also called humanizers) tools," and lists word spinners such as QuillBot as an example of paraphrasing it targets. It won't name which bypasser tools it detects (Turnitin FAQ, Turnitin Guides).

Whether any specific humanizer "works" on any given day is a moving target, and we're not going to help anyone cheat. The more useful point: if a school's policy bans AI-written work, running it through a humanizer doesn't change what it is. If you want help with your own writing, edit in your own voice and keep your drafts.

If Turnitin flagged you

Ask your instructor for the report, look at which passages were highlighted, and gather your version history and drafts. Turnitin's sentence-level false positive rate is around 4% by its own account (Turnitin blog), and some universities, including Vanderbilt, have turned the AI detector off over reliability concerns (Vanderbilt). Our guides on false positives and proving you wrote it walk through what to do.

More guides

Sources

1. Turnitin, "Turnitin's AI writing detection capabilities FAQs" (Turnitin Guides, updated Oct 7, 2026). https://www.turnitin.com/products/features/ai-writing-detection/faq
2. Turnitin, "Using the AI Writing Report" (Turnitin Guides). https://guides.turnitin.com/hc/en-us/articles/22774058814093-Using-the-AI-Writing-Report
3. Turnitin, "Understanding AI writing detection: False positive rates" (Turnitin blog, 2023). https://www.turnitin.com/blog/understanding-the-false-positive-rate-for-sentences-of-our-ai-writing-detection-capability
4. Vanderbilt University Brightspace team, "Guidance on AI Detection and Why We're Disabling Turnitin's AI Detector" (Aug 16, 2023). https://www.vanderbilt.edu/brightspace/2023/08/16/guidance-on-ai-detection-and-why-were-disabling-turnitins-ai-detector/
5. Weber-Wulff et al., "Testing of detection tools for AI-generated text," International Journal for Educational Integrity 19, 26 (2023). https://link.springer.com/article/10.1007/s40979-023-00146-z

This guide is general information, not legal or academic-policy advice. AI detection results, including ours, can be wrong.