Direct answer
Turnitin's AI writing detector does not target one named product; it flags statistical patterns typical of large language models, so text drafted with ChatGPT, Claude, Gemini, and similar assistants can all be detected [1]. It reports the percentage of a document that appears AI-generated rather than naming the model, and it evaluates prose only — not images, diagrams, or code [1]. Because the detector is pattern-based rather than model-specific, no AI tool comes with a built-in exemption, and scores at or below the 20% confidence threshold are shown as *% instead of an exact figure [1].
Which AI Models Can Turnitin's AI Detector Actually Flag?
The short answer is that it is built to flag output from large language models generally, not a single brand. Turnitin describes its detector as trained on patterns characteristic of LLM-generated text, covering GPT-3.5/4-class output as well as Claude- and Gemini-family writing, rather than matching submissions against a stored library of known model responses [2]. That means the question "which AI does Turnitin detect?" is better reframed as "which writing patterns does it detect?" — and those patterns appear across every major assistant.
In practice, the models students ask about most — ChatGPT, Claude, Gemini, Copilot, and similar tools — all produce prose that can trigger the indicator. The detector is looking at how the text is constructed, so a passage written by one model and lightly reworded by another can still register, because the underlying statistical signature survives paraphrasing [2]. There is no "safe" model list, and no assistant is officially exempt.
It is equally important not to read the percentage as a verdict. Turnitin itself advises institutions to treat the AI writing indicator as a signal for further conversation, not as proof of misconduct, and acknowledges that false positives are possible — particularly with highly formulaic academic prose or text that has been heavily edited [2]. A high score is a prompt to review the draft, not an automatic finding.
How Does Turnitin's AI Detector Decide Whether Text Is AI-Generated?
The report does not simply output one number and stop. It shows an overall percentage of the document flagged as AI-generated, alongside a segment-level breakdown that highlights which passages drove the score, so a reader can see exactly where the signal concentrates [3]. This is why two essays with the same headline percentage can look very different once you open the detail view.
The AI writing indicator is also reported separately from the similarity (plagiarism) score, and the two appear side by side in the same report [3]. A draft can have a low similarity score and still carry a high AI indicator, or the reverse — they measure different things and should never be conflated. Instructors see both, which is why "my plagiarism score was fine" is not a defence against an AI flag.
Confidence matters too. When AI detection falls below Turnitin's 20% confidence threshold, the report displays *% rather than an exact percentage, signalling a low-confidence result rather than a precise measurement [3]. Reading segment-level highlights, rather than reacting to the headline number alone, is the most reliable way to understand what the detector actually responded to.
Can I Check My Own Draft for AI Detection Before I Submit It?
Usually you cannot run an institutional Turnitin check yourself. Access to the official checker is controlled by the instructor's assignment settings, so a student cannot normally generate their own report through the university system before the deadline [4]. That gap is precisely why students look for a preview route.
The practical workaround is a preview-only check that does not add your file to the student paper repository. Running a draft through a non-repository check lets you see the AI indicator and the similarity summary before you submit, using the same report type your instructor will see [4]. Because it is the same report format, the score you read in preview is the score you should expect in the institutional system.
Revision then becomes a loop rather than a guess. After you edit flagged passages, re-checking is the only reliable way to confirm the score has actually moved, since the detector responds to the revised text rather than to your intent [4]. Students who treat the preview as a diagnostic — check, revise, re-check — consistently arrive at submission with far fewer surprises.
If you would rather see your own AI indicator before the deadline does, turnitin0 gives you the same report view your instructor will open — an AI writing percentage, flagged segments, and a similarity summary on your draft, delivered in minutes and without adding your file to the repository.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
Get Real Turnitin AI & [Similarity Report](https://www.turnitin0.com/guides/us/turn-it-in-check)
FAQ
Does Turnitin detect ChatGPT specifically?
Turnitin does not single out ChatGPT by name; its detector responds to LLM-style patterns that ChatGPT output commonly contains, so ChatGPT-drafted prose can be flagged [2]. The same applies to other major assistants.
Can Turnitin tell me which AI wrote my paper?
No. The report shows a percentage of text that appears AI-generated and highlights the segments involved, but it does not attribute text to a specific model [1][3].
Is a low AI score the same as a low plagiarism score?
No — the AI writing indicator and the similarity score are reported separately and measure different things, so one can be high while the other is low [3].
What does *% mean in a Turnitin AI report?
It appears when AI detection falls below Turnitin's 20% confidence threshold, indicating a low-confidence signal rather than a precise percentage [1].
Can I see my AI score before submitting?
Not through the institutional checker, which is controlled by your instructor's settings — but a non-repository preview check shows the same report type your instructor will see [4].