Direct answer
Direct Answer - Turnitin checks two distinct things in a student submission: text similarity (how much of your writing matches its repositories of student papers, publications, and internet content) and, when your institution has it enabled, AI writing detection (whether passages look machine-generated). Similarity is reported as a percentage of matched text, not a plagiarism verdict, and the AI indicator is a separate signal with its own confidence threshold. Both appear in the report your instructor sees — and you can preview the same reports before you submit.
What Exactly Does Turnitin Check in a Student Paper — Similarity, AI Writing, or Both?
Turnitin runs more than one check on the same uploaded file, and students often conflate them. The similarity check compares your text against three repository layers: previously submitted student papers, academic publications and journals, and current and archived internet content [1]. That comparison produces a Similarity Report with a percentage and highlighted matches — but that percentage is the proportion of your document that matched something, not a measure of wrongdoing [1].
Separately, Turnitin's AI writing detection evaluates whether the style and structure of your prose looks machine-generated. This is not a plagiarism check at all: text can be 0% similar and still be flagged as AI-written, and text can be heavily matched to sources while showing no AI signal [1]. The two indicators are calculated independently and displayed in different parts of the report.
The practical implication for students is that "what Turnitin checks" is really two questions you must answer about your own draft. A clean similarity score tells you nothing about your AI score, and a low AI score tells you nothing about your matches [1]. You need visibility into both before you submit.
How Does Turnitin's AI Writing Detection Decide Whether Text Is AI-Generated?
Turnitin's AI detector works at the sentence and passage level rather than issuing a single document-wide verdict. It analyzes patterns associated with large language models — predictability of word choice, uniformity of sentence rhythm, and low burstiness — and then reports the share of qualifying text it believes was AI-generated [2]. Crucially, the model only evaluates text that meets a minimum length threshold, so short answers, reference lists, and heavily quoted blocks are excluded from the calculation [2].
Because the output is probabilistic, Turnitin applies a confidence threshold. When the detected AI share falls below roughly 20%, the report displays an asterisk (*%) rather than an exact number, signaling a low-confidence result rather than a confirmed finding [2]. This is why two students with similar drafts can see different-looking reports, and why a "*%" is not the same as "0%."
The detector also has known blind spots that students should understand before they panic. It is designed for English prose, it does not reliably evaluate mathematical notation or code, and it can be fooled by heavily edited or paraphrased machine text [2]. Treat the AI indicator as a directional signal about your drafting process, not as an accusation — and if your score is high, the actionable step is revision, not denial.
Can Students See Their Own Turnitin Similarity and AI Score Before They Submit?
By default, no. Turnitin generates reports for the instructor's assignment inbox, and whether students can view them depends entirely on settings the instructor chooses — many institutions keep student view switched off [3]. That means the first time you see your similarity percentage or AI indicator is often after the deadline, when you can no longer change anything.
Where student view is enabled, the visibility is also delayed. Instructors can release reports immediately, after the due date, or after the assignment is graded, so even a permitted student view may not give you time to act [3]. Students also cannot see the full repository matching detail that instructors see, which makes self-diagnosis harder.
This gap is exactly why pre-submission checking exists. Services such as turnitin0 let you upload a .docx, .pdf, or .txt draft and receive a Turnitin AI detection report plus a similarity report — the same two documents your professor sees in the LMS — in under 15 minutes in 98% of cases [3]. Because the check is non-repository, your file is not added to Turnitin's student paper database and the report is not shared with third-party databases [3]. You get the same information early enough to actually revise.
What Do You Do With the Report Once You Have It?
Reading the report correctly matters as much as getting it. A high similarity percentage is frequently caused by properly cited quotations, a common reference list, or standard academic phrasing — matches that are entirely legitimate and require no change [4]. Instructors are trained to read the highlighted passages in context rather than reacting to the headline number, and you should do the same [4].
If your matches are genuine paraphrasing problems, the fix is citation and rewriting, not cosmetic word-swapping. If your AI indicator is elevated, the fix is substantive revision of the flagged passages so the reasoning and phrasing become your own [4]. Students who understand both signals separately tend to make targeted edits instead of rewriting whole sections blindly [4].
Finally, keep the report. Being able to show your instructor that you checked your draft, understood the flags, and revised accordingly is a strong position to be in if a question ever arises — and it demonstrates the academic honesty the whole system is designed to encourage [4].
If you have just realized you cannot see either number before your deadline, turnitin0 closes that gap: you upload your draft and get the real Turnitin AI and similarity reports back in minutes, so you can revise with the same evidence your instructor will use — instead of guessing.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does Turnitin check for plagiarism and AI at the same time?
Yes — both checks run on the same uploaded file, but they are calculated independently and reported separately [1]. A document can score high on similarity and low on AI, or the reverse.
What does a "*%" AI score mean on a Turnitin report?
It means the detected AI share fell below Turnitin's confidence threshold (around 20%), so the report shows an asterisk instead of an exact percentage [2]. It signals a low-confidence signal, not a confirmed finding.
Will my university let me see my Turnitin report before submitting?
Usually not. Student view depends on instructor settings and is often disabled, and even when enabled the release may be delayed until after the due date [3]. Pre-submission checking is the reliable way to see it early.
Does Turnitin check grammar, spelling, or formatting?
No. Turnitin's core checks are similarity matching and AI writing detection; grammar and spelling are handled by separate tools and are not part of the similarity or AI report [4].
Can Turnitin detect AI text that I edited by hand?
Partially. The detector evaluates statistical patterns in the text, so heavy manual editing can reduce the signal, but results vary and the indicator remains probabilistic [2]. Understanding the report and revising flagged passages is more reliable than assuming edits erased the flag.