Direct answer
Direct Answer - The Turnitin AI detector is a separate, instructor-facing feature that works alongside the similarity report to estimate how much of a submitted document was likely written by an AI tool such as ChatGPT. Its core features include an overall AI percentage, sentence-level highlighting of flagged text, an explicit disclaimer that the score is not proof of misconduct, and a display where any score below 20% appears as an asterisk (*) rather than a precise percentage [1]. The detector targets prose written by mature writers, so it does not reliably assess short passages, non-prose formats, or heavily edited text, and it is designed to support—not replace—human judgment [3].
How Does the Turnitin AI Detector Identify AI-Generated Text and Flag It in a Paper?
Turnitin's AI writing detection evaluates a submitted document in segments and determines the percentage of the text that was likely generated by an AI tool [2]. The report is built around an overall AI score, and it complements that number with sentence-level highlights: sentences the detector believes were AI-generated are marked directly in the text, so instructors can see exactly which passages triggered the flag [2].
The feature is deliberately separate from the plagiarism/similarity check. While the similarity report matches text against a database of sources, the AI writing report analyzes the writing itself for patterns typical of AI generation, which means a paper can have low similarity yet still show a high AI percentage [1]. Instructors access the AI writing report from Turnitin Feedback Studio alongside the similarity report, giving them a combined view of both signals in one place [2].
It is important to note what the feature does not do. The report never declares whether misconduct occurred; it only presents an estimated percentage with the caveat that AI writing in a document does not automatically mean academic dishonesty [1]. Because the model is trained on prose written by mature writers, it is most reliable on long-form academic text and should be treated as an indicator to investigate, not a verdict [2][3].
How Accurate Is the Turnitin AI Detector and How Should the AI Score Be Interpreted?
Turnitin is transparent about the limits of its detector's accuracy. The AI writing detection is optimized for prose written by mature writers, and the fact sheet for educators explicitly warns that it does not reliably detect AI writing in short passages, non-prose formats, or text that has been paraphrased or heavily revised [3]. This means a low or zero score is not a guarantee of originality, and a high score should always be reviewed in context.
The way scores are displayed adds another layer of interpretation. Any overall AI score below 20% is shown as an asterisk (*) rather than as a single-digit percentage, so the only explicit low numeric outcome instructors see is 0%—meaning no AI text was detected [3]. In other words, a paper showing *% is not "AI-free"; it simply falls below the display threshold, which is a nuance many students misunderstand when they preview their results [3].
Turnitin's guidance is consistent: the percentage should never be used as the sole basis for an academic integrity decision [3]. The report is designed as a conversation starter, not a proof tool. Instructors are expected to weigh the AI score alongside the similarity report, the student's writing history, and the assignment context before drawing any conclusion [1][3].
Can Students Run Their Own Turnitin AI Detector Check on a Paper Before Submitting It?
The AI writing report is primarily an instructor-facing feature, so students usually cannot generate the full report on their own drafts inside the institutional system [4]. Whether a student sees the AI writing indicator or any limited report view depends entirely on their university's settings; many institutions leave the report visible only to staff, while the student view of Feedback Studio centers on the similarity report [4].
Because direct access varies by institution and is often unavailable before submission, students who want to preview their AI score need an alternative route [4]. The practical workaround is to run a draft through a service that delivers the same style of Turnitin AI and similarity reports outside the university portal, so the result mirrors what an instructor would see without the student losing access or visibility [4].
Before you submit, you can know exactly what your instructor will see. turnitin0 gives you a real Turnitin AI and similarity report on your own draft—before it ever reaches your university—so you can review your flags, score band, and match highlights in time to revise.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does the Turnitin AI detector also check plagiarism?
No—it is a separate feature. The AI writing report estimates AI-generated text, while the similarity report checks text against a database of sources; instructors see both side by side in Feedback Studio [1][2].
What does *% mean in the AI writing report?
Any overall AI score below 20% is displayed as an asterisk rather than a precise percentage. The only explicit low numeric outcome is 0%, meaning no AI text was detected [3].
Can my instructor's AI report be wrong?
Yes, and Turnitin says so. The detector is optimized for prose written by mature writers and does not reliably detect AI in short passages, non-prose, or heavily revised text, so scores are meant to support human judgment [3].
Will the AI report show which sentences were flagged?
Yes. The report highlights sentences the detector considers AI-generated at the sentence level, alongside the overall percentage [2].
Can I preview my AI score before submitting to my university?
The full report is instructor-facing, and student access depends on institutional settings. To preview the same style of Turnitin AI and similarity results in advance, you can run your draft through a pre-submission check service [4].