Direct answer
Plagiarism detection is the process of comparing a submitted document against large collections of existing text — previously submitted student papers, current and archived web pages, and subscription scholarly databases — and reporting the passages that match. The output is not a verdict but a similarity report: a percentage plus a list of matched sources that a human then interprets. Understanding that pipeline is what lets you predict how your own draft will be read before you submit it [1].
What Databases Does Plagiarism Detection Software Search?
Detection systems are only as wide as the text they can reach, and most institutional systems compare a submission against three distinct pools. The first is a student paper repository: essays, dissertations, and coursework previously submitted at the same or partner institutions. The second is the live and archived internet, which is why a paragraph lifted from an old blog post that no longer exists on the visible web can still surface. The third is a set of subscription scholarly databases covering journals, periodicals, and reference works that ordinary search engines cannot index [2].
Whether your own paper joins that first pool is an institutional setting, not a fixed property of the software. When a repository is enabled, submissions become part of the comparison set for later papers; when it is disabled, the document is checked against existing material but is not retained as a source for future matches. This distinction matters because it determines whether your draft can create a match against a classmate's work or against your own earlier submission [2].
What the system returns is a list of sources, each with a link and a match percentage, plus the specific text that overlapped. Reviewers see which source contributed which share of the matched text, so a single heavily copied paragraph is distinguishable from scattered common phrases drawn from a dozen sources. That structure is deliberate: it hands the judgement back to the reader rather than automating an accusation [2].
How Is a Similarity Score Calculated and What Does a High Percentage Actually Mean?
The similarity index is the proportion of a document's text that matches text found elsewhere. It is a measurement of overlap, not a measurement of misconduct, and the two are frequently confused. A paper with a high index may be quoting extensively with flawless citations, while a paper with a low index may contain a single unattributed passage that is the actual problem [3].
Because the figure is a share of matched text, correct citation does not automatically reduce it. Direct quotations, block quotes, and reference lists all contain text that exists elsewhere, so they register as matches unless the reviewer applies exclusions for quotations, bibliography, or matches below a size threshold. Two reports of the same paper with different exclusion settings can therefore show materially different percentages, which is why comparing your score to a friend's is rarely meaningful [3].
The inverse also holds, and it is the more dangerous error: a low similarity score does not certify originality. Text that has been paraphrased beyond recognition, translated, or generated by a language model may produce few or no matches against a text-comparison database while still failing an integrity review. Similarity analysis answers "does this text already exist somewhere," and it is a separate question from "did the author write this" [3].
Can Students Check Their Own Draft for Plagiarism and AI Writing Before Submitting?
Whether a student can run a self-check inside an institutional account depends on how the institution has configured its settings; self-check and draft folders are not universally available, and where they exist they are usually limited in number or overwritten by later submissions. That variability is why many students look for an independent way to see the same report before the deadline rather than after it [4].
It is also important to keep the two report types separate in your head. Similarity analysis compares your text to existing sources, while AI writing detection estimates how much of the text was machine-generated; the two are produced and reported independently, so a clean similarity score says nothing about the AI indicator and vice versa. A draft that passes one check can still be flagged by the other, and reviewing only one gives a false sense of readiness [4].
A pre-submission check is a diagnostic, not a substitute for the real submission. What it buys you is time: you see the matched sources and the flagged passages while you can still rewrite, add a citation, or restructure a paragraph. Students who check early tend to fix citation gaps and over-quoted sections themselves, which is precisely the outcome a similarity report is designed to prompt [4].
If you would rather see those matched sources and flagged passages on your own draft than discover them in your feedback, turnitin0 runs the same style of check before you submit — an AI detection report and a similarity report delivered together, usually in under 15 minutes, without adding your file to the student paper database.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does a high similarity score mean I plagiarised?
No. The score is the share of your text that matches other sources, and properly cited quotations count toward it unless exclusions are applied [3]. The matched sources, not the number, are what a reviewer actually reads.
Are student papers part of the database?
They can be. Previously submitted work in an enabled institutional repository becomes a comparison source for later submissions, and institutions decide whether that repository is switched on [2].
Why do two checkers give different similarity percentages?
They compare against different databases and apply different exclusion settings for quotations, bibliographies, and small matches. A percentage is only interpretable alongside the source list and settings that produced it [1].
Can I check my own draft before submitting?
Institutional self-check access varies by settings, so availability is not guaranteed [4]. Independent pre-submission checks exist for exactly this reason, and the useful ones return the matched sources rather than a bare number.
Does plagiarism detection also catch AI-written text?
Not through similarity analysis. AI writing detection is a separate indicator, reported separately from the similarity index, so you need to review both before deciding a draft is ready [4].