Direct answer
To check for ChatGPT text, you run the document through a detector trained on large language model output and read the resulting report as a probability signal rather than a verdict. Turnitin's AI writing detection highlights the specific passages it believes were machine-generated and reports the percentage of qualifying prose involved [1]. The practical workflow is: submit the file, read the flagged segments in context, then decide whether the phrasing reflects your own drafting or an AI tool's default style [1].
How Do You Check Whether a Document Contains ChatGPT-Generated Text?
You check a document for ChatGPT text by comparing its sentence-level patterns against a detector trained on output from large language models, including ChatGPT. Turnitin's detector was built specifically to recognise the statistical regularities that LLM prose tends to share — uniform sentence rhythm, predictable transitions, evenly distributed hedging, and low lexical surprise [2]. That is why a detector can flag text you wrote yourself if you happen to draft in a very even, formulaic register.
Scope matters as much as the algorithm. Detection applies to long-form prose above a minimum length, so short answers, bullet lists, reference entries, and heavily quoted material may fall outside what the report actually assesses [2]. A clean score on a 400-word reflection therefore tells you less than you might assume, and a flagged score on a 3,000-word essay tells you more.
Read the output as a probability, not a finding. Turnitin presents the AI score as an indicator that requires human interpretation, and the company is explicit that a high score is a prompt for a conversation rather than proof of misconduct [2]. The useful question is not "did the detector say yes" but "which sentences did it flag, and can I explain how each one was written?"
What Does a Turnitin AI Score Actually Tell You About ChatGPT Text — And What Does the Asterisk Mean?
A Turnitin AI score is the percentage of qualifying prose in your document that the detector attributes to AI generation. It is a proportion, not a grade, and it is computed only over the segments the system considered eligible for detection [3]. A score of 42% means roughly two-fifths of the assessed text showed AI-characteristic patterns — not that 42% of your ideas were unoriginal.
The asterisk is the detail most students misread. When the AI score falls below Turnitin's 20% confidence threshold, the report displays an asterisk — *% — instead of an exact percentage [3]. That asterisk is a low-confidence signal: the detector found something, but not enough to stand behind a number. It is not a zero, and it is not a clean bill of health either. Treat it as "inconclusive, look at the highlighted text."
The report also separates AI-generated text from AI-paraphrased text, and highlights matched sentences directly in the document body so you can see exactly where the signal sits [3]. That distinction matters in practice: a paragraph you drafted and then ran through a paraphrasing tool will be classified differently from a paragraph ChatGPT produced outright, and knowing which is which tells you what to revise.
Can You Check Your Own Draft for ChatGPT Text Before You Submit It?
Yes — and doing so is the single most useful habit in this whole process. Turnitin frames AI detection as a conversation starter between educators and students rather than an enforcement verdict, which only works if students have seen the same signal their instructor will see [4]. Reviewing your own draft against the criteria your instructor applies, before the deadline rather than after, turns a stressful result into an editing task.
Start with your institution's assignment brief. Institutions are advised to state their expectations for AI use explicitly, so students know in advance what will be flagged and what counts as permitted assistance [4]. If the brief is silent, ask before you submit — not after a report lands in your inbox.
Then check the draft yourself and read the flagged passages line by line. Where the detector points at a sentence, ask whether you can reconstruct how it was written. If the answer is "I prompted for it," you have a revision job; if the answer is "I wrote it, but I write in a very even register," you have a note to add to your process — and a reason to vary sentence length and add specific, sourced detail in your own voice.
Reading the flags is the easy part; the harder part is getting a real report on your own file before your instructor does. That is exactly what turnitin0 was built for — an independent pre-submission check that returns the same Turnitin AI detection report and similarity report your professor sees in the LMS, so you are reviewing the actual signal rather than guessing at it. Files are checked without being added to Turnitin's student paper database, English documents between 300 and 30,000 words are accepted, and most orders finish within 5–15 minutes.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does Turnitin detect ChatGPT text specifically, or all AI writing?
Turnitin's detector is trained on patterns characteristic of large language models including ChatGPT, so it is not limited to one product [2]. Text drafted in Claude or Gemini can be flagged by the same mechanism.
What does an asterisk instead of a percentage mean on my report?
An asterisk (*%) appears when the AI score falls below Turnitin's 20% confidence threshold, meaning the detector found a signal but not enough to report a number [3]. It signals low confidence, not a clean result.
Can a human-written paragraph be flagged as ChatGPT text?
Yes. Detection responds to statistical regularity in prose, so very even, formulaic academic writing can register as AI-characteristic [2]. That is why the report highlights specific sentences for human review rather than issuing a verdict.
How much of my document does the AI score actually cover?
The score is a percentage of the qualifying prose the detector assessed, not of the whole file [3]. Short passages, lists, and quoted material may sit outside detection scope entirely.
Should I check my draft before submitting it?
Reviewing your own draft against the same criteria your instructor applies is the recommended approach, since AI detection is intended as a starting point for discussion rather than an enforcement outcome [4].