Direct answer
A plagiarism checker's similarity score is the percentage of your document's text that matches sources in the checker's database — it is a measure of overlap, not a verdict of plagiarism. A 22% score can be entirely innocent (quoted material, a reference list, common phrasing), while a 9% score can contain one unattributed copied paragraph that matters far more. Turnitin's own guidance frames the number as a starting point for reading the report source by source, which is why the score alone never tells you whether your draft is safe to submit [1].
What Does the Similarity Score in a Plagiarism Checker Actually Mean?
The similarity score is a ratio: the words in your submission that the system matched to an existing source, divided by the total words checked, expressed as a percentage. It is a total of all matches combined, so a single source cited in four places contributes to one cumulative figure rather than four separate ones [2]. That arithmetic detail is the reason two papers with the same score can be in completely different situations — one is padded with a reference list, the other has a copied methodology section.
Matches are also grouped and reported per source, which is where the real information lives. Turnitin's report lists each matching source with its own percentage and lets you open the matched passages in context, so you can see instantly whether the overlap sits in your bibliography, a block quotation, or your own argument [2]. Report settings let you exclude quoted material, bibliography, and small matches below a chosen word count; toggling those exclusions on and off is often the fastest way to understand what is actually driving a number.
A high score driven by properly cited references is a formatting issue, not an integrity issue, and most institutions treat it that way. The number that deserves attention is the one that survives after you exclude quotes and references — that residual percentage represents text that appears in someone else's work without visible attribution from you [1][2]. Reading the report in that order (score, then sources, then residual) turns a stressful number into an actionable checklist.
What Similarity Percentage Is Considered Too High, and How Do Universities Use It?
There is no universal cutoff. Institutions set their own thresholds in academic-integrity policy, and many deliberately avoid publishing a fixed number precisely because a percentage cannot distinguish poor paraphrasing from deliberate copying [3]. Some departments flag submissions above a stated level for manual review; others ask markers to review the report for every submission regardless of score. If you want the rule that applies to you, it is in your module handbook or the university's academic-integrity page, not in a forum thread.
What markers actually do with the report is more consistent than the thresholds suggest. They open the matched passages and judge whether the borrowed text is attributed, whether quotation marks enclose direct wording, and whether the paraphrase is genuinely rewritten rather than synonym-swapped [3]. A 30% score composed of cited sources and a standard methods description usually passes without comment; a 12% score containing one uncited paragraph lifted from a journal article usually does not.
Practically, scores in the very high range (roughly above 40–50%) most often signal a structural problem rather than a citation slip: a reference list counted in full, a template or cover sheet shared across a cohort, or a large copied block [3]. Very low single-digit scores are not automatically a clean bill of health either, since a short submission or a checker with a narrow database can under-report. Treat the percentage as a diagnostic signal and the source list as the evidence.
How Can I Check My Own Similarity Score and AI Score Before I Submit?
Previewing your own report before submission converts guesswork into a fixable list. A pre-submission check returns the same two documents your institution sees: a similarity report with per-source matches, and a separate AI writing report. Those two scores are independent measurements and are not interchangeable — a low similarity score says nothing about whether your prose reads as machine-generated, and a clean AI score says nothing about unattributed copying [4].
The AI score has its own reporting convention worth knowing before you panic. Turnitin displays an asterisk (*%) instead of an exact percentage when its AI detection confidence falls below its 20% threshold, meaning the signal is low-confidence rather than zero [4]. Students regularly misread that asterisk as a failure or as a perfect result; it is neither. Reading both reports side by side, before the deadline, is the only way to know which passages need a citation added and which need rewriting in your own voice.
The workflow that works is deliberately unglamorous: finish the draft, run the check, open the similarity report, exclude quotes and references, and read every remaining highlighted passage in context. Then decide whether each one needs a citation, a genuine paraphrase, or deletion [4]. Doing this at least a day before the deadline leaves room to revise rather than to hope.
Reading a score is only useful if the score is the one your marker will see. turnitin0.com returns the actual Turnitin AI and similarity reports — the same PDFs that land in your institution's system — so you can open the source list, check the flagged passages, and fix them while there is still time to revise.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is a 20% similarity score bad?
Not on its own. Twenty percent is a common, unremarkable figure when a paper includes a reference list and quoted sources; what matters is whether the matched text is attributed. Open the report, exclude quotes and bibliography, and look at what remains.
Why is my similarity score higher than my friend's when we wrote different essays?
Database coverage and reference-list length drive scores as much as writing does. A paper citing heavily from sources in the checker's index will score higher than one citing books and primary data outside it [1].
Does a low similarity score mean my work is original?
It means little text matched the checker's database — not that the argument is your own. Short submissions, unusual topics, and sources outside the index can all suppress the number [2].
Can the similarity score and the AI score be the same number?
No. They are separate reports measuring different things: text overlap with existing sources versus patterns associated with AI generation [4]. A submission can score low on one and high on the other.
Should I check before I submit, or is that cheating?
Previewing your own draft to verify citations is a revision step, and institutions generally expect students to submit original, correctly attributed work. What matters is that you act on the report by fixing attribution rather than trying to disguise overlap.