Direct answer
A plagiarism score detector compares your document against a database of web pages, publications, and previously submitted student papers, then reports the percentage of your text that overlaps with those sources. That percentage is a similarity index, not a verdict: it measures matched text, and it cannot by itself tell you whether you plagiarised. Turnitin's own guidance states that the Similarity Report is designed to be interpreted source by source, because a high score can come entirely from correctly cited quotations and a bibliography [1]. What matters for your submission is not the single number but which sources matched, how much of each match is quoted or cited, and whether the checker you used searches the same database your institution does [1].
What Does a Plagiarism Score Actually Measure, and What Similarity Percentage Is Considered Safe?
A plagiarism score measures overlapping text, and nothing more. When a detector processes your draft, it breaks the document into fragments, searches its index for near-identical strings, and adds up the length of the matching fragments relative to your total word count. Turnitin's report separates those matches into internet sources, publications, and student papers, and each match is colour-coded so you can open it and read the exact overlapping passage [2]. That granularity is the point: the headline index is a summary of a list of sources, and the list is where the actual information lives [2].
There is no universal "safe" percentage, and any checker that advertises one is oversimplifying. A 22% score built from six properly cited quotations plus a reference list is a different document from a 22% score built from two uncited paragraphs lifted from a journal article, even though the number is identical. Turnitin's documentation is explicit that matches must be read in context, because quoted material and bibliographies legitimately inflate the index [1]. A reference-heavy literature review can sit well above a bare percentage that would be alarming in a short reflective essay.
The practical rule is to treat the score as a triage signal, then audit the matches. Open every source in the report and ask three questions: is this passage quoted and cited, is it a common phrase or a title, or is it uncredited borrowing that needs rewriting? Turnitin also lets you exclude quotes and the bibliography, which changes the index and gives you a cleaner view of uncited overlap [2]. Because that setting alters the number, you should never compare your score against a friend's unless you know both reports used identical exclusions [2].
Finally, remember what the score cannot see. It cannot evaluate whether your paraphrase preserves meaning, whether your citation style is correct, or whether you used an AI writing tool. Those are separate judgements, which is why institutions pair similarity reports with AI writing detection and with human marking. A clean similarity index is a necessary condition for a confident submission, not a sufficient one [1].
Why Does the Same Essay Get a Different Plagiarism Score on Different Detectors?
The dominant reason is the comparison database, not the algorithm. A detector can only match text it has already indexed, so a checker without a repository of previously submitted student papers will systematically miss the largest source of accidental overlap in coursework — recycled essays, shared drafts, and material from earlier cohorts [3]. Turnitin's help documentation makes this dependency explicit: the score is a function of what the system holds, so the same paragraph can score 4% on one platform and 19% on another without either being wrong [3].
The second reason is tokenisation and matching thresholds. Systems differ in how many consecutive words they require before flagging a match, whether they normalise punctuation and citation markers, and how they handle quoted text. A checker that strips quotations before scoring will report a materially lower index on a quotation-heavy essay than one that counts everything [3]. This is why a low score from a free tool is weak evidence and a high score from a strict tool is not automatically an accusation [3].
The third reason is timing. Web content changes, new papers enter repositories, and indexes refresh on different schedules, so a score is a snapshot of one moment against one index. Turnitin's guidance on reading the report stresses that the source list, not the percentage, is the stable evidence you can act on [3]. Two reports that disagree are giving you two different views of your text; both can be internally accurate.
The practical consequence is that only one score matters for your grade: the one produced by the system your institution actually runs. If your university uses Turnitin, a check on a different engine — however reputable — is an approximation. That is the reasoning behind checking your draft against the same engine, on the same document format, before the deadline rather than after it [3].
How Can I Check My Own Plagiarism Score Before I Submit My Assignment?
Start by checking the draft you will actually submit, in the format you will actually upload. A .docx file preserves headings, tables, and reference lists that a plain-text paste destroys, and those structures affect how matches are counted. Turnitin's student-facing guidance recommends reviewing the report source by source after the check, because the value of a pre-submission run is the list of matches you can still fix [4].
Second, use a check that does not deposit your file into a student paper database. If a pre-submission checker adds your draft to the repository it searches, you can create the very match you were trying to avoid when a classmate checks a similar draft later. A non-repository check returns the report without enrolling your document as a future source, which is the appropriate mode for a draft you intend to revise [4]. Turnitin's own advice to students is to focus on improving attribution and paraphrase rather than chasing a target number, and that only works if the report you receive is diagnostic [4].
Third, read the report the way your marker will. Work down the match list in descending order of overlap, open each source, and classify it: correctly quoted and cited, a generic phrase that needs no action, or uncredited borrowing that must be rewritten in your own words with a citation. Turnitin's guidance is that this source-level review, not the headline percentage, is what determines whether a submission is sound [4]. Keep a short list of the passages you rewrote so you can confirm they are clean on a second run.
Fourth, budget time for a re-check. Fixing matches changes the document, and a revised draft deserves a fresh report before the deadline. Students who check once, rewrite heavily, and submit without re-running the detector are working from stale evidence. Two runs — one diagnostic, one confirmatory — is the minimum sensible sequence, and it is the pattern Turnitin's student guidance implicitly supports by treating the report as a revision tool rather than a final verdict [4].
If you want that confirmatory run to come from the same engine your marker uses, turnitin0 exists for exactly that step: upload your .docx, .pdf, or .txt draft and receive the Turnitin AI detection report and the similarity report as two downloadable PDFs in a single order, showing the score, the flagged passages, and the match highlights your professor would see. Checks are non-repository, most orders finish in 5–15 minutes, and 100,000+ reports have been delivered to 20,000+ students. A single check is $3.80, with a 10-check pack at $2.75 each and no subscription.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is a 0% plagiarism score good?
It means no matching text was found in that detector's index, which is not the same as a well-argued or correctly cited essay. A 0% score can also reflect a checker with a thin database [3].
Does a high similarity score mean I will be accused of plagiarism?
No. Turnitin's guidance is that the report is interpreted in context, and correctly quoted, cited material legitimately raises the index [1]. Review each match before drawing conclusions.
Can my professor see that I used a checker before submitting?
Not if the check is non-repository. A repository check enrols your file as a potential future match, while a non-repository check returns the report without adding your document to the student paper database [4].
Why does my score change between two runs on the same draft?
Indexes refresh, so new web pages and papers can match on a later run. Check that both reports used the same exclusions for quotes and bibliography before comparing [2].
Should I aim for a specific percentage before submitting?
Aim for a report with no unexplained matches. Turnitin's student guidance is to fix attribution and paraphrase rather than chase a number, because the source list is the actionable evidence [4].