Direct answer
A GPT detector such as Turnitin's AI writing detection does flag ChatGPT-written text. Turnitin segments a submission into sentences and marks the ones its model predicts were generated by an LLM, then reports the share of qualifying text that was flagged [1]. Crucially, a displayed *% means the AI score sits below Turnitin's 20% confidence threshold, so the report is signalling a low-confidence result rather than an exact percentage [1]. In practice, whether your ChatGPT draft gets flagged depends on how heavily you leaned on the model and how much you rewrote afterwards.
How Does a GPT Detector Like Turnitin's AI Detector Identify ChatGPT-Written Text?
A GPT detector does not read your essay the way a human marker does. Turnitin's model works at sentence level: it splits the submitted document into sentences, evaluates each one, and highlights the sentences it classifies as likely AI-generated, then aggregates those into an overall percentage of qualifying text [2]. That sentence-level design is why a report can show a low overall score while still highlighting a handful of individual sentences.
The underlying signal is statistical predictability rather than a fingerprint hidden in the words. LLM output tends to be more uniform in word choice and rhythm than human drafting, and the detector is trained to recognise those patterns across text produced by GPT-3.5, GPT-4 and comparable models [1]. This is also why the model is version-sensitive: Turnitin states that its detection is tuned to specific model families, and results on newer or heavily edited text can be less reliable [1].
Two practical caveats matter for anyone interpreting a report. First, false positives are possible, and a flagged sentence is a signal to review, not automatic proof of misconduct [2]. Second, only text that meets a minimum length threshold is scored at all, so short answers or heavily quoted passages may not contribute to the percentage [2].
Can Turnitin's GPT Detector Actually Catch ChatGPT Output, and What Score Does It Give?
Yes — Turnitin's detector is explicitly built to identify text from ChatGPT and similar large language models, and it will return a percentage for submissions that contain enough qualifying text [1]. The score you see is the proportion of the document's qualifying text that the model flagged, not a verdict on the whole essay, which is why two students with similar ChatGPT use can receive very different numbers [2].
Reading the number correctly is the part most students get wrong. When Turnitin displays *%, the AI score is below its 20% confidence threshold — the report is telling you the signal is weak, not that the text is clean or that it is 100% AI [1]. A high, specific percentage, by contrast, means a large share of qualifying sentences were classified as AI-generated [2].
Whether you can check your own draft before the real submission depends on your institution. Turnitin's student-facing guidance is that draft checking is governed by instructor and institutional settings, and students should ask their instructor whether a pre-submission check is permitted [3]. Many students therefore use an independent pre-submission check to see a Turnitin AI report on their own draft before the graded upload happens [3].
How Can You Lower the Turnitin AI Score on ChatGPT Text Before You Submit?
Start with disclosure and revision, not concealment. Turnitin's own guidance to institutions is that students are generally expected to disclose AI assistance, and that revising flagged passages in your own voice is the recommended response — detection output is designed to prompt a conversation, not to serve as standalone proof of misconduct [4]. Editing that genuinely changes sentence structure, adds your own analysis and varies rhythm does more than swapping synonyms [4].
Where a draft was generated almost entirely by ChatGPT, surface-level edits often leave the statistical pattern intact. The more reliable route is rewriting the flagged passages so the reasoning, examples and phrasing are yours, then re-checking the result rather than assuming the edit worked [4]. Because Turnitin's report is sentence-level [2], you can target the specific highlighted sentences instead of rewriting the whole document.
That is also why re-checking matters: a rewrite can shift the score, and the only way to know is to see a fresh report on the revised draft. Turnitin0 is built for exactly this loop — an independent pre-submission check that returns a Turnitin AI detection report and a similarity report together, so you can compare the flagged sentences before and after you revise [3].
If rewriting ChatGPT text by hand still leaves sentences highlighted, turnitin0's AI humanizer is the faster path: it rewrites flagged passages while preserving your meaning, citations, headings and.docx formatting, and it is built for text drafted with ChatGPT, Claude or Gemini.
※ Turnitin0.com - AI Humanizer Clearing All AI Flag of ChatGPT Text
FAQ
Does Turnitin's GPT detector flag every ChatGPT essay?
No. The detector scores only qualifying text above a minimum length threshold and highlights individual sentences, so a heavily edited ChatGPT draft can return a low or *% result [1][2]. A clean-looking score reflects a weak signal, not a guarantee that no sentence was flagged.
What does *% mean on a Turnitin AI report?
*% is shown when the AI score falls below Turnitin's 20% confidence threshold, meaning the detector's signal is not strong enough to report a precise figure [1]. Treat it as low confidence rather than as a zero.
Can I check my ChatGPT draft against Turnitin before I submit it?
That depends on your instructor and institution, since draft checking is controlled by local settings [3]. An independent pre-submission check returns the same style of AI and similarity reports so you can review flagged sentences in advance [3].
Is a flagged sentence proof that I used ChatGPT?
No. Turnitin's detection output is intended to support academic judgement and can produce false positives, which is why the recommended response is to revise flagged passages in your own voice and discuss the result with your instructor [2][4].
Will rewriting ChatGPT text in my own words lower the score?
Substantive rewriting — changing structure, adding your own analysis and varying sentence rhythm — is the response Turnitin recommends, and re-checking the revised draft is the only way to confirm the effect [4]. Humanizing tools exist for the same purpose: turnitin0's AI humanizer rewrites flagged passages while keeping meaning, citations and formatting intact.