Direct answer
A plagiarism detection system is software that compares a submitted document against a large indexed corpus of other texts and reports the passages that overlap, expressed as a similarity percentage rather than a verdict on intent. Turnitin's own documentation is explicit that the similarity score represents the proportion of a document that matches other sources, and that interpreting those matches is a human judgment made by the instructor [1]. In practice, the system does three things: it ingests your file, it searches for matching text across its reference collections, and it returns an annotated report showing exactly which passages matched, where they came from, and how much of your paper they account for [1]. Understanding that pipeline is what lets you read a similarity report calmly instead of panicking at a number.
How does a plagiarism detection system actually work?
At its core the engine is a matching machine, not a reader. It breaks your submission into text sequences, creates a representation of those sequences, and then looks for the same sequences in its indexed reference content — it is comparing strings and word order, not evaluating whether you understood your sources [2]. That is why the system reacts to long runs of identical wording and stays quiet when you genuinely rewrite an idea in your own sentence structure, and it is why a properly quoted and cited block can still light up as a match [2].
Because the comparison is sequence-based, the report you receive is a list of matched passages with source attribution attached to each one, and it deliberately stops short of declaring misconduct. The engine surfaces evidence; the instructor, tutor, or editor decides whether that evidence represents poor citation practice, acceptable quotation, or something more serious [2]. This division of labour matters when you are reading your own report, because a high percentage and a plagiarism finding are not the same thing.
It also explains the behaviour people find most confusing: two papers on the same topic can produce very different scores depending on how much conventional phrasing they share. Standard academic connectors, definitions of well-known concepts, and reference list entries are all text, and text is what the engine matches [2]. A system that flagged nothing at all on a literature-heavy essay would arguably be the suspicious one.
What databases does a plagiarism detection system compare your paper against?
Turnitin's similarity report organises every match into three distinct source categories: internet pages, periodicals and publications, and previously submitted student papers [3]. The internet bucket covers live and archived web content, the publications bucket covers journals, books, and other periodical material, and the student paper bucket covers work that has been submitted to the system before — which is the category most students underestimate, because it is the one that catches resubmitted or shared assignments [3].
That third bucket is also the one your institution controls. Administrators decide whether submissions are added to the student paper repository, and it is entirely possible to run a check in a mode where your file is compared against the existing corpus but is not itself retained in it [3]. This is a meaningful distinction if you are checking a draft, sharing work with classmates, or reusing material from an earlier course.
The report itself tells you how the score was assembled rather than just giving you a number. Each match is listed with its source and the percentage of your document it contributes, so a single heavily-quoted source can account for most of your total while dozens of tiny matches add up to very little [3]. Reading the breakdown is usually more informative than reading the headline percentage, because it tells you whether you have one problem or twenty.
Can you check your own paper with the same plagiarism detection system before you submit it?
Usually not through your institution. Student access to Turnitin's checking tools is provisioned by the instructor or administrator, which means a student cannot simply log in and run an unlimited number of checks on their own drafts [4]. If your course has not enabled a draft or self-check assignment, the institutional route is closed to you regardless of how early you start.
That gap is why independent pre-submission checking exists. Students who want to see the report before the graded submission typically use a third-party service that returns the same style of similarity and AI writing reports, so the feedback arrives while there is still time to act on it [4]. The practical value is timing: a check run three days before the deadline is a revision tool, whereas a check run after submission is just information you cannot use.
And the fixes a pre-check reveals are usually small and mechanical rather than dramatic. Unquoted overlap, paraphrasing that stays too close to the original wording, and missing in-text citations are the recurring culprits, and all three are correctable in an afternoon once you can see the flagged passages in context [4]. Checking early converts an anxious unknown into an ordinary editing task.
Reading about how the engine matches text is useful, but it is a poor substitute for seeing your own passages highlighted with their sources attached. turnitin0 gives you that view before the deadline, returning a similarity report and an AI writing report on your draft so you can fix the overlap while it still counts for something.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is a high similarity score the same as being found guilty of plagiarism?
No. The score tells you what proportion of your document matched other sources; it does not assess intent or whether the overlap was properly cited [1]. Instructors read the underlying matches before drawing any conclusion.
Why does my reference list show up as matched text?
Because the engine matches text sequences, and a bibliography is a list of titles and author names that also appear in its indexed sources [2]. Most departments exclude or discount reference sections when they review a report.
Does checking my paper add it to the student paper repository?
It depends on how the check is configured. Institutions decide whether submissions are retained, and non-repository checks exist that compare your file without adding it to the corpus [3].
Can I run the institutional check myself as many times as I like?
Generally no — access is set up by your instructor or administrator, so self-service checking is not something students can switch on themselves [4]. Pre-submission checks through an independent service are the common workaround.
What should I fix first if my draft is flagged?
Start with unquoted overlap and paraphrasing that tracks the original too closely, then check for missing in-text citations [4]. Those three account for most correctable matches.