Direct answer
Turnitin's AI detector — officially called AI writing detection — is a model built into the Similarity Report that flags text likely generated or modified by AI tools such as ChatGPT [1]. When a paper is processed, the detector produces an overall AI writing percentage and highlights the segments it believes an AI tool wrote or rewrote [1]. The indicator is visible to instructors and administrators, not to students [1]. This guide explains how the detector works, what the score really means, and how you can check your own draft before submitting.
How Does Turnitin's AI Detector Work?
The AI Writing Report sits alongside the Similarity Report and opens with an overall percentage of "qualifying text" the model predicts could have been generated by an AI tool or further modified by an AI paraphraser or bypasser [2]. Turnitin highlights those predicted segments directly in the submission, with an interactive breakdown bar that lets instructors jump from a highlight to the exact page [2]. Crucially, the AI percentage is independent of the similarity score, so a low plagiarism match says nothing about AI generation [2].
Under the hood, Turnitin extracts sentences and splits them into overlapping segments, then scores each segment between 0 and 1 for how likely it is to be AI-generated; overlapping sentence scores are pooled and aggregated into the document-level percentage [1]. The logic rests on word probability: large language models pick the next most probable word very consistently, while human writing is far more idiosyncratic, and the classifiers are trained to measure that difference [1].
Only prose sentences in a long-form document count as qualifying text — poetry, scripts, code, bullet points, tables, and annotated bibliographies are not reliably detected [2]. The English detector also catches text that was likely AI-generated and then modified with an AI paraphraser or "humanizer" tool, while the Spanish and Japanese detectors do not include that capability [2].
Detection currently covers major models including GPT-4o through GPT-5.x, Gemini, Claude, LLaMA, Mistral, DeepSeek, and Grok, and in July 2026 Turnitin consolidated its multi-model ensemble into a single model while keeping the false positive rate below 1% [1]. Submissions must contain at least 300 words of prose in a supported language (English, Spanish, Japanese, or Arabic) to generate a report at all [2].
What Does a Turnitin AI Detection Score Mean?
The percentage is simply the share of qualifying prose that the model believes may have been written or modified by an AI tool — it is not a measure of plagiarism, and it is not a verdict of misconduct [2]. Because the number is easy to misinterpret, Turnitin's official guidance hub devotes separate articles to it, including "How should I review the AI Writing report?" and "What should I do if the AI writing percentage is high?" [3].
Turnitin deliberately hides scores between 1% and 19%, displaying them as % instead of a number [2]. The reason is transparency: false positives occur more often in the 0–19% band, so an asterisk signals that the low score is less reliable [2]. A report can show 0%, a % bucket, or a numeric percentage from 20% to 100%, and instructors who see a high percentage are pointed to official guidance on how to review it [3].
Turnitin states the model is designed to keep the false positive rate under 1% and that the indicator should never be used as the sole basis for an academic-integrity decision [1]. The score is one piece of evidence that instructors weigh alongside their own judgment and institutional policy [1]. In other words, a high AI percentage starts a conversation; it does not end one [3].
How Can I Check My Paper With Turnitin's AI Detector Before Submitting?
Here is the reality most students discover too late: you cannot self-check a paper inside Turnitin without uploading it to an official assignment created by your instructor [4]. The AI writing indicator itself is only visible to instructors and administrators, and there is no public portal where a student can run the official detector on demand [1].
Turnitin's official options are narrow. If your institution enables Turnitin Draft Coach, you can run your own similarity, citation, and grammar checks inside Google Docs or Microsoft Word before submission [4]. Otherwise, your only path is an assignment that allows resubmissions: the first three attempts generate reports immediately, while a fourth resubmission forces a 24-hour wait, and the daily resubmission limit is three [4].
If your instructor has disabled resubmissions, your first upload is final and you get no preview at all [4]. Even the resubmission route mainly generates similarity reports — it does not give you a freely readable AI writing report the way instructors get one [2].
That is why students who want to see their real Turnitin AI percentage and similarity result before submitting turn to a service that runs the official Turnitin detector on their draft and returns the actual reports. Doing this before the deadline turns an anxious unknown into a number you can act on — and, if the AI score is high, a problem you can still fix.
Because Turnitin doesn't give students a free self-check, the practical way to see your real AI percentage before submission is to run your draft through the official detector first. turnitin0 does exactly that: upload your.docx,.pdf, or.txt file and receive the actual Turnitin AI writing report together with the similarity report — the same reports your instructor sees — delivered within minutes, with no subscription, no archival of your paper, and nothing sent to any third-party database.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Can students see Turnitin's AI detection score on their own paper?
The AI writing indicator is shown to instructors and administrators; student visibility depends entirely on institutional settings and on whether the instructor shares the report [1].
What does *% mean on a Turnitin AI report?
A star means the report found some AI signal but below the 20% threshold; Turnitin hides single-digit percentages because false positives are more likely in the 0–19% band [2].
Does Turnitin's detector flag Grammarly or basic grammar fixes?
The detector targets text generated or rewritten by large language models and AI paraphrasing tools; ordinary grammar-correction edits are not what the model is trained to flag [1].
Can I check my paper for AI before I submit it?
Officially, students can only self-check via Turnitin Draft Coach (if enabled) or an assignment that allows resubmissions [4]. A practical alternative is checking with a service that returns the official Turnitin reports, such as turnitin0, before your final submission.
Is a high AI percentage proof of academic misconduct?
No. Turnitin explicitly states the indicator should not be used as the sole basis for action; instructors are expected to apply human judgment and institutional policy [1][2].