Direct answer
AI copy text is text produced by a generative model — ChatGPT, Claude, Gemini — that you copy out of the chat window and paste into a document, and the practical question that follows is almost always the same: will this text be flagged when it is checked? Turnitin's AI writing detection runs inside the Similarity Report on the file you actually submit, so the score you receive is tied to the document, not to the chat transcript it came from [1]. That distinction matters more than most students expect, because the same paragraph can score differently depending on how much of it you rewrote, how it is formatted, and how much qualifying prose surrounds it. Understanding what the detector reads — and what it ignores — is the difference between guessing and knowing before your deadline.
What Does Turnitin's AI Detector Actually Look At When It Scores Copied AI Text?
Turnitin's AI writing detection does not read your chat history, your prompt, or your drafting process. It analyses the submitted document and estimates what share of its qualifying prose was likely generated by an AI model, then presents that estimate as an overall percentage inside the Similarity Report your instructor already opens [2]. The unit of analysis is the sentence, not the essay, which is why a report can show a modest overall figure while still highlighting long stretches of flagged text.
The phrase "qualifying prose" carries real weight. Reference lists, bibliographies, and quoted material are handled differently from your own body paragraphs, so a document padded with citations is not scored as if every word were your argument [2]. This is one reason two students with near-identical AI copy text can receive very different reports: the proportion of the document that counts toward the score is not the same.
Turnitin is also explicit that the indicator is probabilistic. A flagged sentence is a signal that the wording resembles model-generated prose, not a finding that you did not write it, and the company positions the report as one input alongside assignment context and drafting history rather than a verdict on its own [2]. That framing is worth internalising, because it is the same framing your tutor is trained to use when the report lands in their marking queue.
Why Does AI Copy Text Sometimes Score 0% AI and Sometimes Show an Asterisk Instead?
The asterisk is the single most misread symbol in the whole report. Turnitin displays *% in place of an exact figure when the detected AI share falls below its 20% confidence threshold, which means the system is not confident enough in the signal to publish a number [3]. Students routinely read *% as "zero" and relax, or read it as "caught" and panic — both are wrong. It is an inconclusive low-confidence reading, and Turnitin's own guidance tells instructors to treat it that way rather than as evidence of authorship either way [3].
The reason the same essay can move between *% and a hard number is that detection responds to surface texture as much as to substance. Heavily edited AI copy text — sentences broken up, clauses reordered, vocabulary swapped, examples inserted — stops resembling the statistical pattern the model is looking for, so the flagged share can drop below the threshold and collapse into an asterisk [3]. Conversely, clean, uniform, well-behaved academic prose of the kind models produce by default tends to hold a visible score.
There is a second-order effect that catches people out: a document that is mostly your own writing with one AI-generated section can produce a lower overall percentage than the section deserves, because the flagged sentences are diluted by qualifying human prose around them [3]. Never read the headline figure without opening the sentence-level view — the number tells you how much was flagged, the highlighting tells you where.
Can You Check AI Copy Text With a Real Turnitin Report Before You Submit?
Yes — and for most students the useful move is to see the report before the deadline, not after. Turnitin's published guidance stresses that the AI indicator is meant to be interpreted in context, and that context is exactly what you have and your instructor does not: your notes, your drafts, your revision history [4]. Seeing the same report view your tutor will see converts a guessing game into a concrete editing decision — which sentences to rewrite, which to attribute, which to leave alone because they are already your own words.
What you are looking for when you check is not a single pass/fail number. Read the overall percentage, then read the sentence-level highlighting, then read the similarity matches separately, because plagiarism and AI detection answer different questions about the same file [4]. A document can carry a clean similarity score and a flagged AI section, or the reverse. Treating them as one metric is how students end up "fixing" the wrong problem the night before submission.
The practical constraint is that institutional Turnitin access usually does not let you self-submit a draft, and where it does, the draft may enter the paper repository. That is the gap an independent pre-submission check fills: a non-repository run that returns the AI detection report and the similarity report as two downloadable PDFs, matching the report format instructors see, typically in under fifteen minutes. For a student holding a document full of AI copy text, that is the difference between submitting on evidence and submitting on hope [4].
If you want that certainty before your deadline rather than after it, turnitin0 gives you the report view itself — the AI percentage, the sentence-level flags, and the similarity matches — on your own draft, without adding your file to the student paper database.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does copying AI text into a Word document change how Turnitin scores it?
Formatting alone does not change the detection result, but what you do on the way in does. Editing, restructuring, and adding your own examples alters the prose the detector reads, which is why a revised draft can score differently from the raw chat output [2].
Is *% on a Turnitin AI report the same as 0%?
No. The asterisk appears when the detected AI share is below Turnitin's 20% confidence threshold, so it signals a low-confidence reading rather than a confirmed zero [3]. Check the sentence-level view before drawing conclusions.
Can I lower an AI score just by paraphrasing copied text?
Paraphrasing can reduce the flagged share, but it does not add your own analysis, and the detector responds to more than vocabulary swaps. Rewriting the reasoning in your own words and grounding it in your sources is what actually changes both the report and the quality of the work [4].
Does Turnitin detect AI in reference lists and quotes?
Turnitin distinguishes qualifying prose from reference material and quoted passages when computing the score, so citations are not treated as your own generated text [1]. Keep referencing accurate and complete; it helps the report read correctly.
What should I do if my draft shows a high AI percentage?
Open the sentence-level highlighting and work through the flagged passages one by one, rewriting in your own voice and adding source-specific detail [4]. Re-check the draft afterwards to confirm the change registered before you submit.