Direct answer
Turnitin checks a submitted document for two separate things: text similarity (how much of your writing matches its repositories of student papers, web pages, and subscription publications) and AI-generated writing (which segments its model classifies as machine-written). It does not check for plagiarism in the legal sense, does not grade your argument, and does not scan for spelling, citation style, or factual accuracy — those remain the instructor's job. The similarity percentage and the AI indicator are independent scores, and both are reported to your instructor alongside the matched passages themselves [1].
What does Turnitin's similarity report actually compare your paper against?
The similarity report compares your text against three distinct repositories that Turnitin maintains: a student paper repository, web content indexed from the live internet and archived pages, and a subscription publications index covering journals, periodicals, and books [2]. Each repository contributes matches independently, and the report lists them as numbered sources with their own individual match percentages rather than collapsing everything into one opaque figure [1]. That structure matters because a 24% overall similarity could be one heavily quoted paragraph or twenty scattered four-word overlaps — the number alone does not tell you which.
Whether your own paper is added to the student repository depends on an institutional setting your instructor controls, not on anything you choose at upload time [2]. When that setting is on, your submission can later be matched against other students' work, including work submitted at other institutions that use the same system. Instructors can also exclude quotes, the bibliography, and matches below a set word count, all of which change the percentage you see without changing the underlying text [2].
The critical framing is that similarity is a measurement, not a verdict. Turnitin's own documentation is explicit that the percentage indicates the proportion of a document matching its sources and that interpretation belongs to the instructor, who can see the context of every highlighted passage [1]. A high similarity score from a correctly cited literature review and a high similarity score from copied prose look identical in the number but completely different in the report body.
How does Turnitin decide which parts of a paper are AI-generated?
Turnitin's AI writing indicator works at the segment level, not the word level. The system splits the document into passages, runs each through a model trained on both human-written and AI-written academic text, and reports the proportion of the document its model classifies as AI-generated [3]. Because the output is a probability per segment, the report highlights specific spans of text rather than declaring an entire essay "AI" or "not AI."
A detail that surprises most students: when the analysis does not meet Turnitin's confidence threshold, the report does not print a precise figure at all — it displays an asterisk (*%) to signal a low-confidence result [3]. Seeing *% is not the same as seeing 0%, and it is not the same as seeing a confirmed detection. It means the model could not make a confident call either way.
The AI indicator is also architecturally separate from the similarity score [3]. A passage can be flagged as AI-generated while showing zero similarity, and a passage can show high similarity to a published source while carrying no AI flag at all. These are two different questions answered by two different systems, which is why a clean similarity score is not evidence that your AI score is clean — or the reverse [1].
How can a student see their own AI and similarity report before submitting?
The practical answer is that you check your own draft first, because the report your instructor receives is interpretation-dependent and you cannot see it from the outside. Turnitin frames AI detection as a conversation aid for instructors rather than an automatic misconduct finding, and it encourages students to understand what the report shows and to discuss flagged sections openly [4]. That conversation goes far better when you already know what your draft looks like on the report.
Institutional access also varies. Not every student sees the same report view, and the AI indicator is not universally enabled across every assignment [4]. Two students in the same cohort can therefore have very different levels of visibility into their own results, which makes a pre-submission preview the only reliable way to know where you actually stand.
This is where turnitin0 fits. You upload your .docx, .pdf, or .txt and receive two downloadable PDFs in one checkout — a Turnitin AI detection report and a similarity report — showing the same scores, flags, and match highlights your professor sees in their LMS. English documents only, over 300 and under 30,000 words, under 20 MB. Turnaround is under 15 minutes in 98% of cases, and the check is non-repository, so your file is not added to Turnitin's student paper database and your report is not shared with third-party databases. Over 100,000 reports delivered to 20,000+ students, with a 4.9/5.0 satisfaction rating.
You have just seen what Turnitin actually measures — similarity against three repositories, an AI indicator that can return *% when it is unsure, and an instructor who reads both in context. The one thing you cannot do is read those two numbers off your own draft without running it. turnitin0 gives you that view before the deadline does, in the same format your instructor will open.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does Turnitin check for plagiarism, or just similarity?
It measures similarity — the proportion of your text matching its repositories — and leaves the plagiarism judgment to your instructor [1]. The report gives them the matched passages and sources; the conclusion is human.
What does *% mean on a Turnitin AI report?
It means the AI analysis did not reach Turnitin's confidence threshold, so no precise percentage is shown [3]. It is a low-confidence signal, not a confirmed detection and not a clean zero.
Does Turnitin check grammar, citation style, or factual accuracy?
No. Turnitin's checked dimensions are text similarity and AI-generated writing [1][3]. Spelling, referencing format, argument quality, and factual correctness are assessed by your instructor, not by the report.
Can I see my AI and similarity report before I submit?
Not through your institution, because access to the AI indicator varies by assignment and institution [4]. A pre-submission check through a service like turnitin0 returns both PDFs so you can review the scores and flags first.
Will Turnitin flag my own previous work as similar?
It can. If your institution has the student paper repository enabled, previously submitted papers can match against new submissions, including your own earlier work [2]. Instructors can exclude small matches or bibliography entries, which changes the visible percentage.