Direct answer
To detect similarity, you run a document through a checker that compares its text against three bodies of material: previously submitted student papers, live and archived web pages, and licensed academic databases and journals. The checker highlights every matched passage and reports the total matched text as a percentage [1]. Detecting similarity is therefore a two-part job: producing the comparison, then reading the highlighted matches in context rather than trusting the headline number [1].
What Does a Similarity Checker Actually Compare?
A similarity checker does not scan "the internet" as a single undifferentiated blob. Turnitin's comparison set is built from a student paper repository of work previously submitted to institutions, a web crawler that captures current and archived pages, and a set of licensed academic databases and journal collections [2]. Each of those three pools behaves differently. The repository catches recycled essays from earlier cohorts, the web index catches copied blog and study-site prose, and the journal databases catch lifted academic phrasing that a free web search would never surface [2].
This is why two checkers can return wildly different numbers for the same file. If a tool only searches the open web, it is blind to the repository and to paywalled journals, which is precisely where a lot of undergraduate copying originates [2]. When you compare tools, compare their comparison sets first and their interface second.
The report structure matters just as much as the score. Turnitin lists every matched source separately with its own percentage contribution, so a 24% total might be one 20% source plus four 1% fragments, or twelve 2% sources — two completely different problems [2]. Exclusion settings for quotations, bibliography, and matches below a word threshold also change the displayed figure, which is why the same draft can show different percentages under different settings [2]. Always read the source list before reacting to the total.
What Similarity Percentage Is Acceptable in a University Submission?
There is no universal acceptable similarity percentage. Turnitin states plainly that no fixed threshold exists and that institutions and instructors set their own expectations, often varying by discipline, assignment type, and level of study [3]. A literature review that quotes heavily and a short reflective piece can legitimately produce very different totals, so a single "safe number" is a myth worth discarding early.
The more important correction is that the score is not a verdict. A low percentage does not prove originality — a fully paraphrased argument built from one uncredited source can score low while still being plagiarised — and a high percentage does not prove misconduct [3]. Turnitin frames the report as a tool for interpretation, not a detector of cheating, and instructors are expected to open the matches and judge them [3].
In practice, most students worry about the band between roughly 10% and 25%. Treat that band as a prompt to investigate rather than a pass/fail line: open each match, ask whether the highlighted text is a correctly attributed quotation, a properly cited paraphrase, or an unattributed lift, and act on the third category only [3]. If your department publishes a numeric limit, that limit governs; if it does not, the quality of your attribution governs.
Can You Check Similarity Yourself Before Submitting?
Yes — and doing it before submission is the single highest-value habit in this whole process. Students can run a draft through a checker that returns a full report in advance, which turns a blind submission into a review step where you still have time to fix what the report flags [4]. The workflow is straightforward: check the draft, open every highlighted match, then either cite it, quote it properly, or rewrite it in your own words [4].
One technical detail changes the outcome substantially. If you check a draft through a repository-based system, that draft may be stored and later match against itself when your final version is submitted, producing a false positive you then have to explain [4]. A non-repository check avoids that trap entirely, because the file is compared but not added to the student paper database. When you are self-checking, confirm the repository behaviour before you upload, not after.
Finally, keep the report rather than just the number. The source list tells you which passages need work, and re-checking after revision tells you whether the fix actually landed [4]. Two checks — one diagnostic, one confirmatory — are worth more than a dozen glances at a percentage.
Reading your own matches is only useful if the report you are reading is the one your instructor will see. turnitin0 runs your .docx, .pdf, or .txt through a non-repository check and returns the same two PDFs an institutional system produces — a similarity report with highlighted matches and source-by-source percentages, plus an AI writing report — usually in under 15 minutes, with 100,000+ reports delivered to 20,000+ students and a 4.9/5.0 satisfaction rating. Nothing is added to Turnitin's student paper database, and you can delete your files afterwards.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does a similarity score tell me whether I plagiarised?
No. The score is the percentage of text that matched a source, and Turnitin explicitly frames the report as something to interpret rather than a judgement [1]. Only a human reading the highlighted passages can decide whether a match is a cited quotation, a weak paraphrase, or an unattributed lift [3].
Why does my similarity score change between two checks of the same file?
Because the comparison set and the exclusion settings both shift. Repository content grows as new papers are submitted, the web index updates continuously, and settings for quotations, bibliography, and small-match thresholds alter which passages are counted [2].
Can I detect similarity without adding my paper to a database?
Yes. A non-repository check compares your document against the same kinds of sources but does not store it in the student paper database, so your final submission will not match against your own draft [4].
Is a 0% similarity score a good result?
Not automatically. A low score can reflect heavy paraphrase of a single uncredited source just as easily as genuine originality, which is why the source list matters more than the total [3].
How many sources should appear in a normal report?
There is no expected count. A handful of small matches from common phrasing is routine, while one large single-source match is usually the one worth examining first [2].