Direct answer
Turnitin is a text-matching and academic-integrity service that institutions license so submitted work can be compared against a large indexed corpus — student papers, web pages, and publication databases — and returned to the instructor as a similarity report with matched passages highlighted [1]. It also runs a separate AI writing detection model that estimates how much of a document was likely generated by an AI tool [3]. Nothing is "read" by a person at submission time: the system produces scores, source lists, and highlights, and a human interprets them [2].
What does Turnitin actually check when you submit a paper?
Turnitin is not a single plagiarism scanner; it is a comparison engine that runs your file against three distinct corpora. The first is Turnitin's own repository of previously submitted student papers, the second is indexed web content, and the third is a set of subscription publication databases covering journals, periodicals, and books [2]. Your document is fingerprinted and matched against all three, which is why a paragraph you wrote yourself can still surface a match if a similar phrase exists somewhere in that corpus [1].
The output is a similarity report, not a verdict. The similarity score is the percentage of your submission that matched something in the corpus, and it deliberately includes legitimate matches such as quoted material, common phrasing, and bibliography entries [1]. Instructors are trained to read the score together with the highlighted passages and source list rather than treating the number as a decision on its own [2].
Institutional settings materially change what the report shows. Instructors can exclude quotes, exclude the bibliography, exclude small matches below a word threshold, and control whether the submission is stored in the student repository [2]. Two students submitting identical text under different assignment settings can therefore see different scores, which is a common source of confusion when classmates compare results.
The practical takeaway is that Turnitin answers a narrow question — "how much of this text overlaps with indexed sources?" — and leaves the judgement to the instructor. Understanding that boundary is the first step to reading your own report correctly [1][2].
How does Turnitin's AI writing detection differ from its similarity check?
AI writing detection is a completely separate model from similarity matching, and it answers a different question: how much of this document looks like it was produced by a generative AI tool [3]. Instead of matching text against a corpus, the model analyses patterns in the writing and returns a percentage estimate along with a segment-level breakdown showing which sentences or paragraphs contributed to the signal [3]. A submission can carry a high similarity score from properly cited sources while showing no AI signal at all, and a fully original AI-generated draft can show a strong AI signal with a low similarity score [3].
Confidence thresholds matter more than most students realise. Turnitin only displays an exact AI percentage when the model's confidence is high enough; when the signal falls below that threshold, the report shows an asterisk (*%) instead of a precise number [3]. An asterisk is therefore a low-confidence indicator, not a hidden or withheld score, and it should not be read as a clean zero.
Because the two systems are independent, the reports are also delivered and interpreted separately. The similarity report is about overlap with existing sources; the AI report is about stylistic probability [3]. Instructors who understand this distinction look at both, and they weigh the AI indicator alongside the highlighted segments rather than in isolation [2][3].
For students, the operational consequence is simple: a low similarity score does not automatically mean a low AI score, and vice versa. If you want to know where you actually stand on both axes before a deadline, you need to see both reports rather than assuming one predicts the other [3].
How can students see their Turnitin similarity and AI report before submitting?
A Turnitin report is a standardised artefact: a similarity percentage, a colour-coded source list, and clickable inline highlights that show exactly which passage matched which source [4]. The same layout is used whether the report is opened by an instructor in the institutional system or generated outside a class assignment, which is what makes a pre-submission preview genuinely comparable to the instructor view [4]. Reports can also be downloaded as PDF, so a preview can be kept, re-read, and compared line by line against the final submission [4].
The catch is that institutional settings still shape the report. Because quote, bibliography, and small-match exclusions are configured at the assignment level, a preview run outside the class repository will not be byte-identical to the version your instructor sees [2][4]. What a preview does give you is the same underlying detection output — the matches, the flagged segments, and the AI signal — so you can see which passages are exposed and revise them before the real deadline submission [1][4].
This is why many students check their draft first and submit afterwards. Seeing the report early turns an opaque percentage into a concrete list of passages to rework, and it removes the guesswork from deciding whether a paragraph needs rephrasing or a citation [4]. It is also the only reliable way to know whether the AI indicator applies to your draft at all, given how the confidence threshold works [3].
If you want that preview without waiting on an instructor to run it for you, a pre-submission check that returns the same style of AI and similarity PDFs is the practical route — and that is exactly what the next section covers [4].
Reading the mechanics is one thing; seeing your own numbers before the deadline is another. turnitin0 lets you upload your draft and get back the same two instructor-facing PDFs — the AI detection report and the similarity report — so the flags, matches, and score bands stop being theoretical and become something you can act on while you still have time to revise.
※ Turnitin0.com - Actual [Turnitin AI Report](https://www.turnitin0.com/) Cover, Score, Flag And Similarity Summary
FAQ
Is Turnitin the same thing as a plagiarism checker?
Not exactly. Turnitin is a text-matching service that compares your submission against student papers, web content, and publication databases, then returns a similarity score with highlighted matches [1][2]. The similarity score is an overlap percentage, not a plagiarism ruling — an instructor still has to interpret the highlighted passages and decide what they mean [1].
Can Turnitin tell the difference between human and AI writing?
It runs a separate AI writing detection model that estimates the share of a document likely generated by an AI tool, reported as a percentage with a segment-level breakdown [3]. When the model's confidence is low, the report shows an asterisk (*%) rather than an exact figure, so the indicator should be read as a probability signal, not proof [3].
Why do two students get different similarity scores for the same text?
Because instructors configure assignment settings — excluding quotes, excluding the bibliography, ignoring small matches, and controlling repository storage — and those settings change what the report displays [2][4]. The underlying matches are the same; the presentation and exclusions are not [2].
Do I get an AI score even if I wrote everything myself?
Yes, the AI model runs independently of the similarity check, so a fully human draft can still receive an AI indicator if its style triggers the model [3]. Many students preview their draft before submitting precisely so they can see whether the AI signal applies and which passages contributed to it [4].
Can I see my Turnitin report before I submit my final version?
Yes. A pre-submission check returns the same style of AI and similarity PDFs your instructor would see, so you can review the flagged segments, matches, and score bands while you still have time to revise [4]. Just note that assignment-level exclusions may still differ between a preview and the institutional run [2][4].