Direct answer
Turnitin does not license a third-party tool: it uses its own proprietary AI writing detection model, developed in-house and informed by a research collaboration with OpenAI, then embedded directly into its Similarity platform [1]. Instead of scanning for a single "AI signature," the model analyzes academic prose sentence by sentence and flags segments that appear to have been generated by AI tools such as ChatGPT rather than written by a person [1]. When AI-written text is found, the instructor-facing AI writing report shows an overall percentage with the suspect passages highlighted so the evidence can be reviewed at segment level [1].
How Does Turnitin's Own AI Writing Detection Technology Work?
Turnitin's detector is a custom machine-learning model, not an off-the-shelf classifier bought from another vendor. When Turnitin announced the capability in April 2023, it explained that the model was built on Turnitin's proprietary technology and refined with OpenAI, drawing on a large corpus of academic writing together with AI-generated text so the system learns the statistical patterns of machine-produced prose [2]. Because the training data centers on student-style academic writing, the detector is tuned for essays, reports, and research papers rather than for casual or conversational content [2].
Once a paper is submitted, the detector evaluates the document in segments and identifies which sentences or paragraphs appear to have been AI-generated [2]. It produces an overall AI score for the submission and highlights flagged segments directly in the writing report, letting an instructor see precisely which parts of the document triggered the indicator instead of receiving a single unexplained verdict [2].
The model is designed to look at prose only. Quoted material, bibliographies, and non-prose elements such as tables and bullet lists are excluded from the analysis so that legitimate citations do not inflate the result [2]. This segment-level design is why the same document can show a high AI score in one passage and a clean result in another, giving reviewers a transparent, evidence-based starting point for their own judgment [2].
How Accurate Is Turnitin's AI Detector at Spotting ChatGPT and Other AI Writing?
Turnitin reports that its detector is highly precise on the long-form academic prose it was built to assess, with internal testing showing a false-positive rate of under 1% for documents that are almost entirely AI-written and around 2% when less than 20% of a document is AI-generated [3]. In practice this means a fully human-written paper is very unlikely to be flagged as AI, while clearly machine-generated submissions are caught at a high rate [3].
Accuracy, however, is not uniform across every input. Detection is strongest on longer documents written in fluent English, and it weakens when text is short, heavily edited, translated, or rephrased by another tool, because those transformations erode the statistical patterns the model was trained to recognize [3]. Turnitin has been transparent that no detector can guarantee a perfect verdict on every sample of text [3].
For this reason, Turnitin positions the report as a signal for educators to weigh alongside their own knowledge of the student and the assignment, not as the sole basis for an academic-integrity decision [3]. Instructors are advised to review the flagged segments themselves, discuss the findings with the student, and avoid relying on the percentage alone, which is exactly why the report exposes highlights rather than a single black-box judgment [3].
Can Students Preview Turnitin's AI Detection Results Before Submitting Their Paper?
In most university setups, the AI writing report is visible to the instructor inside the Turnitin grading interface, while students typically do not see an AI percentage or the flagged passages from their own account before submission [4]. The Similarity report may be configured for student preview at an institution's discretion, but the AI writing report is generally treated as a staff-facing integrity tool rather than a student self-check [4].
Turnitin's own documentation frames the AI writing report as part of the instructor's review workflow: it is meant to be opened alongside the similarity score when marking a submission, and the flagged segments are the evidence an educator evaluates before deciding how to respond [4]. That workflow assumption means students who want an early warning usually cannot obtain one from the same account they submit through [4].
As a result, students who rely on AI assistance and want to know what their draft will look like to Turnitin before the official submission need an independent way to see the real report. Previewing the exact AI score, flags, and similarity summary on the final version of the paper — before it is sent to the university — turns the detector from a surprise into a manageable, reviewable result.
That is exactly the gap turnitin0 was built to close. turnitin0 gives students the actual Turnitin AI and similarity reports on their own draft before they submit, showing the same AI percentage, segment flags, and similarity breakdown their instructor would see in the institutional system. Upload the finished paper, review exactly how Turnitin's proprietary detector reads it, and decide with full information whether to revise, rewrite, or submit with confidence.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
1. Is Turnitin's AI detector the same as OpenAI's ChatGPT classifier?
No. Turnitin built its own proprietary detection model and collaborated with OpenAI during development, but the system running inside Similarity is Turnitin's model, not OpenAI's retired text classifier, and it is trained specifically on academic writing [2].
2. Can Turnitin detect AI text from Claude, Gemini, and other AI tools?
Turnitin's detector is designed to identify text produced by large language models, including ChatGPT-era tools, though performance is strongest on fluent, long-form prose [3]. Text that has been heavily rewritten, translated, or generated by newer or less common models may be harder to classify with certainty.
3. What does a *% AI score in a Turnitin report mean?
In Turnitin's AI writing report, any score below 20% is displayed as *% rather than as a single-digit number, so a *% result simply means the detector found little or no AI-written text [1]. Only a confirmed 0% appears as an explicit low numeric outcome.
4. Does Turnitin flag AI text even if I paraphrase it?
Sometimes. The detector scores segments of prose, and light editing usually does not hide machine-written passages, but substantial rewriting or translation can reduce detectability because the original statistical patterns are altered [3]. Turnitin advises treating any score as a signal that requires human review.
5. Why can't I see the AI writing report in my student Turnitin account?
The AI writing report is normally released to instructors as part of the marking workflow, and student accounts are rarely given access to it before submission [4]. Students who want an early read on their draft typically need to check it through a service that returns the real Turnitin reports before they submit.