Direct answer
Turnitin's AI generator checker is a built-in feature of the Similarity Report that estimates how much of a submitted document was produced by an AI writing tool such as ChatGPT, Claude, or Gemini. It splits the text into sentence-level segments, marks the ones its model considers AI-generated, and returns an overall percentage of AI-written prose alongside the highlighted passages [1]. The score is a probability signal, not proof of authorship, and Turnitin displays it as an asterisk (*%) whenever the detected share falls below its 20% confidence threshold [1].
What Does Turnitin's AI Generator Checker Actually Detect?
Turnitin's AI writing indicator does not look for a hidden watermark or a file signature left behind by ChatGPT. It analyses the statistical behaviour of your prose — how predictable each next word is, how uniform sentence rhythm and vocabulary are — and compares those patterns against a model trained on both human and AI-written academic text [2]. Text that is highly predictable and evenly paced tends to score higher, which is why polished, formulaic paragraphs are flagged more often than messy, idiosyncratic ones [2].
The detection runs only over continuous prose. Reference lists, quotations, and some structured elements are excluded from the calculation, so a report that flags 30% of a document is describing 30% of the prose segments, not 30% of the total word count [1]. The report also colour-codes flagged sentences, which matters far more than the headline number: a single flagged paragraph in a 4,000-word essay and a scattered flag across every section are very different problems [1].
Importantly, the checker reports a pattern, not an author. Turnitin itself frames the indicator as a signal that should prompt a conversation or a closer look, not as a determination that a student cheated [2]. That distinction is the reason the same report can be read very differently by two markers — and the reason students benefit from seeing their own report before the deadline rather than after it.
How Accurate Is Turnitin's AI Detection and What Do the Score Bands Mean?
Turnitin publishes a false-positive rate for fully human-written submissions of roughly 1% or less at the 20% threshold, meaning the large majority of purely human text is not flagged at that level [3]. The number is a rate, not a guarantee: it describes aggregate behaviour across millions of submissions, so an individual document can still be flagged despite being written entirely by hand [3]. That residual risk is why the low-confidence band exists in the first place.
The score bands are the part students misread most often. Anything under 20% is displayed as *% because Turnitin considers the signal too weak to report as a firm percentage — it is a low-confidence flag, not a clean bill of health [1]. From 20% upward the report shows an actual figure, and the highlighted segments let you see exactly which sentences drove it [1]. A 100% score means every analysed prose segment was classified as AI-generated, which usually indicates a document pasted wholesale from a chatbot [2].
Accuracy also depends on what you feed it. Short documents, heavily quoted passages, technical writing with repeated terminology, and non-native English prose that has been heavily edited can all produce noisier results [3]. Turnitin advises institutions to treat the indicator as one piece of evidence among several, which is a useful frame for students too: the score tells you where a marker will look, not what a marker will conclude [3].
How Can Students Check an AI-Generated Draft Before Submitting It?
The safest workflow is to see the report before your institution does. Institutional Turnitin usually does not let students run an unlimited number of self-checks on the same assignment, and repeated submissions can be logged against your account, so many students use an independent pre-submission check to preview the AI and similarity results first [4]. Turnitin0 is built for exactly that: you upload a .docx, .pdf, or .txt file (English, over 300 words, under 30,000 words, under 20 MB) and receive two downloadable PDFs in one order — a Turnitin AI detection report and a similarity report matching what instructors see in their LMS. Checks are non-repository, so your file is not added to Turnitin's student paper database, and turnaround is under 15 minutes in 98% of cases.
Reading the preview correctly matters as much as getting it. Start with the flagged segments rather than the percentage, and ask of each one whether the phrasing is genuinely yours or whether it is the smooth, generic cadence that AI drafting tends to leave behind [4]. If a paragraph was drafted with ChatGPT, Claude, or Gemini and then lightly edited, that paragraph is the one most likely to be flagged — and it is also the one where rewriting in your own voice does the most good [2]. Students who fix the flagged passages rather than the headline number consistently get the cleanest re-check.
Keep the timing realistic. Run the check when the draft is substantively finished, leave room for a revision pass, then re-check the revised file so you can confirm the flagged share actually dropped [4]. If you are revising under time pressure and the flagged text is extensive, the AI humanizer service on turnitin0 rewrites flagged passages while preserving meaning, citations, headings, and .docx formatting, and it is designed for text drafted with ChatGPT, Claude, or Gemini — lowering the Turnitin AI score to *% or even 0%, or your money back.
A preview only helps if it is the real thing — the same report format, the same score display, and the same flagged passages your marker will open. That is what turnitin0 delivers: an actual Turnitin AI and similarity report on your own draft, in minutes, before the deadline decides the outcome for you.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does Turnitin's AI checker detect ChatGPT text reliably?
It detects it often enough that ChatGPT-drafted prose is the most common source of high scores, but reliability varies with how heavily the text was edited and how formulaic it is [2]. Treat the score as a strong hint about where a marker will look, not a certainty [3].
What does an asterisk (*%) instead of a number mean?
It means the detected AI share fell below Turnitin's 20% confidence threshold, so the report withholds an exact figure [1]. It is a low-confidence signal rather than a confirmed pass, and heavily edited text can sit in that band.
Can I check my own draft in Turnitin before I submit it?
Institutional accounts usually restrict self-checks and log repeat submissions, so many students preview results through an independent checker first [4]. Turnitin0 returns a Turnitin AI report plus a similarity report on your own file without adding it to the student paper database.
Will paraphrasing flagged sentences lower the score?
Usually yes, if you genuinely restructure the sentence rather than swapping synonyms, because the detector responds to predictability and rhythm rather than specific words [2]. Re-check the revised file to confirm the flagged share dropped [4].
Does Turnitin's AI checker look at reference lists and quotes?
No — the calculation targets continuous prose, so quotations and reference lists are excluded from the flagged share [1]. That is why a high score can coexist with a bibliography that is entirely your own work.