Direct answer
The Turnitin AI detector is a feature inside Turnitin's similarity checking system that estimates how much of a submitted paper's prose was likely generated by an AI writing tool, and it reports that estimate as a percentage of qualifying text [1]. It is a separate signal from the similarity (plagiarism) score, so a paper can show 0% similarity and still carry a high AI indicator [1]. Turnitin presents the result as an advisory indicator rather than a verdict, and its own guidance tells instructors to interpret it alongside knowledge of the student and the assignment [1].
How does the Turnitin AI detector decide which parts of a paper are AI-generated?
The detector does not look for "AI phrases" from a blacklist. It evaluates how predictable and how uniform the writing is across sequences of words, which is why it reacts to the flat, evenly-paced rhythm that large language models tend to produce [2]. Text that varies sentence length, takes unexpected turns, and uses specific, situated detail tends to score low, while templated, generic, evenly-structured prose tends to score high [2].
Detection runs on prose only. Quotations, references, and bibliographies are excluded from the analysis, and short passages below the minimum analysed length are not scored at all — which is why a heavily quoted literature review can look "clean" while the discussion section carries all the risk [2]. The report then highlights the specific segments that were flagged, so the reader sees which paragraphs drove the number rather than a single opaque figure [2].
Two practical consequences follow. First, the indicator is probabilistic: Turnitin states that its model can produce false positives, particularly on highly formulaic writing such as lab reports, standard essay structures, and text written by non-native English speakers working from templates [2]. Second, because the analysis is segment-based, a low overall score does not guarantee that no individual paragraph was flagged — the segment view is where the real detail lives [2].
Why does a Turnitin AI report show an asterisk instead of a percentage?
An asterisk in place of a number is not an error and it does not mean zero. Turnitin displays the asterisk when the overall AI score falls below its 20% confidence threshold, signalling that the detection signal is too weak to state a reliable figure [3]. In practice, an asterisk means the report found low-confidence AI indicators rather than a confident finding either way [3].
This threshold exists because the underlying model outputs a probability, and small probabilities are not meaningful enough to publish as a precise percentage. Reporting "*%" prevents a student or an instructor from treating a 3% reading as hard evidence of anything, while still leaving the underlying segment data visible for anyone who wants to inspect it [3].
The distinction matters at submission time. If your report shows an asterisk, the correct reading is "no confident AI signal was detected at the document level," not "this paper is certified AI-free" [3]. Institutional policies vary, and some markers will still open the segment view and read the highlighted passages themselves, so the asterisk is a starting point for a conversation rather than a clean bill of health [3].
Can you check your own Turnitin AI score before you submit?
Usually not through your institution. Students generally cannot run their own submission through the university's Turnitin integration, because the assignment is created and controlled by the instructor, and many courses do not enable draft-check attempts at all [4]. Where a draft-check assignment does exist, it is the safest route, since it uses exactly the same configuration your final submission will face [4].
When no draft check is available, students typically turn to third-party pre-submission services that return a Turnitin-style AI and similarity report on a draft [4]. These services are independent and not affiliated with Turnitin, so the honest expectation is a close preview of the two signals — the AI indicator and the similarity matches — rather than a guaranteed identical score [4]. What they are genuinely useful for is timing: catching a flagged discussion section, or an over-quoted paragraph, while there is still time to rewrite it [4].
The workflow that works is straightforward. Finish the draft, run a pre-submission check, read the segment view rather than only the headline number, and revise the specific flagged passages in your own voice — adding concrete detail, varying sentence length, and replacing generic transitions [4]. Then re-check. Repeating that loop is what actually moves a score, and it is far more reliable than guessing which sentences "sound like AI" [4].
If you would rather see the real thing than guess, turnitin0 runs the same two reports your instructor sees — the AI writing indicator with its segment highlighting, plus the similarity report — on your own draft in under 15 minutes in 98% of cases. That way the asterisk, the percentage, and the flagged paragraphs are all in front of you while the work is still yours to fix.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is the Turnitin AI detector the same as the similarity score?
No. The similarity score measures overlap with existing sources, while the AI indicator estimates how likely the prose was machine-generated [1]. They are reported separately and can move in opposite directions.
Does an asterisk mean my paper scored 0% AI?
It means the overall score fell below Turnitin's 20% confidence threshold, so no reliable figure was displayed [3]. Treat it as a low-confidence signal, not a guarantee.
Can a human-written essay be flagged as AI?
Yes. Turnitin acknowledges false positives are possible, especially with formulaic structures, heavily templated writing, or text produced under heavy editing constraints [2].
Which parts of my paper are actually analysed?
Prose only — quotations, references, and bibliographies are excluded, and passages below the minimum analysed length are not scored [2]. The flagged segments are highlighted in the report.
Can I see my AI score before I submit?
Only if your instructor enables a draft-check assignment [4]. Otherwise students use independent pre-submission checkers, which are not affiliated with Turnitin and provide a close preview rather than an identical score [4].