Direct answer
An AI detector is software that estimates how much of a piece of writing was generated by a large language model rather than written by a human. In academic settings the detector most students actually meet is Turnitin's AI writing detection, which runs inside the similarity report and returns a percentage of qualifying text flagged as AI-generated [1]. That percentage is an estimate built from sentence-level classification, not a verdict, and it is designed to give an instructor a signal to review rather than a conclusion to act on [1].
What Does a Turnitin AI Detector Actually Measure, and What Does Its Score Mean?
Turnitin's AI detector measures the proportion of a document's qualifying prose that its model classifies as AI-generated, and it reports that as a percentage alongside a sentence-level breakdown [2]. Only sentences with enough words are analysed, which is why short answers, headings, tables, and bullet lists often sit outside the calculation entirely [2]. The number you see is therefore a measure of flagged text coverage, not a measure of how much of your thinking came from a model.
The score needs careful reading because Turnitin deliberately hides precision at the low end. When the detected amount falls below the model's 20% confidence threshold, the report displays an asterisk-percentage marker (*%) instead of an exact figure, signalling a low-confidence result rather than a precise low score [2]. Students frequently misread that marker as a hidden or secret number, when it actually means the detector is not confident enough to commit to a value [2].
It also helps to separate the two reports that arrive together. The similarity report compares your text against a database of sources and highlights matching passages, while the AI writing report estimates authorship style; they answer different questions and a high similarity score does not imply a high AI score [2]. Reading them side by side is the only way to understand what an instructor will see.
Finally, the score is explicitly an estimate rather than proof of misconduct. Turnitin positions the indicator as one input for a human judgement, which is why institutional policies normally require an instructor to interpret the report in context before raising any concern [2].
Why Do AI Detector Results Differ Between Tools, and How Reliable Are They?
Different detectors disagree because they are not measuring the same thing in the same way. Each tool uses its own trained model, its own confidence threshold, and its own rules for splitting text into analysable units, so two detectors can return very different numbers for the same paragraph [3]. Treating one tool's score as a universal truth is the single most common mistake students make.
Reliability also depends heavily on the text itself. Turnitin's detection is tuned for long-form academic prose, and its accuracy shifts with genre, language background, editing history, and how much a draft has been revised by hand [3]. A polished human essay and a lightly edited AI draft can look statistically similar, which is exactly why the company publishes caveats about false positives rather than claiming certainty [3].
There is a second, subtler source of disagreement: the confidence threshold. A detector that reports *% below 20% and a detector that reports a raw 12% may be describing the same underlying signal with different display conventions [3]. Comparing those two numbers directly, or comparing a free checker's output to an institutional report, produces confusion rather than clarity.
The practical takeaway is to treat any single detector score as one signal among several. If a result surprises you, the useful response is to look at which specific sentences were flagged and why, not to chase a lower number from an unrelated tool [3].
How Can Students Preview Their Turnitin AI and Similarity Reports Before Submitting?
In most institutions, students cannot generate their own official Turnitin reports before submission. The check is normally configured so that only the instructor's assignment folder produces a report, and students either see nothing or see the report only after the deadline has passed [4]. That is a deliberate design choice, not an oversight, and it means there is no legitimate student-side button inside your university account.
Because of that gap, a pre-submission preview has to come from outside the institutional system. Independent checking services let a student upload a .docx, .pdf, or .txt file and receive a Turnitin AI detection report plus a similarity report in the same checkout, formatted the way an instructor would see it [4]. The important technical detail is whether the service is non-repository: a check that does not add your file to Turnitin's student paper database avoids the awkward situation where your own draft later matches itself in the official submission [4].
Timing matters as much as the report itself. A preview is only useful if it arrives while you can still revise, which is why turnaround of minutes rather than days is the practical requirement for a last-minute deadline [4]. It is also worth remembering that a preview is a diagnostic, not a guarantee: it tells you where the detector is likely to flag you so you can rewrite those passages in your own voice [4].
Turnitin0 is an independent service that is not affiliated with Turnitin, LLC, and it exists precisely to fill this pre-submission gap for students in the US, UK, Canada, Australia, New Zealand, and Ireland. It has delivered over 100,000 Turnitin AI and similarity reports to more than 20,000 students, with an average turnaround under 15 minutes and a 4.9/5.0 satisfaction rating; a single check costs $3.80, and a 10-check pack works out to $2.75 per check with no subscription [4].
If you want to know exactly what your instructor's report will say before the deadline closes, turnitin0 lets you see the same AI score, flagged sentences, and similarity summary in minutes — so you can fix the passages that matter instead of guessing.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is an AI detector the same as a plagiarism checker?
No. A plagiarism checker compares your text against source databases and highlights matches, while an AI detector estimates whether the writing style looks machine-generated [1]. Turnitin runs both inside one similarity report, but they answer different questions and should be read separately [2].
What does the *% symbol on a Turnitin AI report mean?
It means the detected AI-generated portion fell below Turnitin's 20% confidence threshold, so the tool shows a marker instead of an exact percentage [2]. It is a low-confidence signal, not a hidden high score.
Can I run an official Turnitin check on my own draft?
Usually not. Institutional settings typically restrict report generation to the instructor's assignment folder, so students cannot produce their own official report before submitting [4]. Independent non-repository preview services exist for exactly this reason [4].
Why did two detectors give me completely different scores?
Because each tool uses a different model, threshold, and text-segmentation rule, so the outputs are not directly comparable [3]. Genre, language background, and how much you edited the draft also shift the result [3].
If I am flagged, does that mean I cheated?
No. Turnitin presents the AI indicator as an estimate for instructor review, and institutional policies normally require a human to interpret the report in context before drawing any conclusion [1][3].