Turnitin similarity is a metric generated by the Turnitin plagiarism detection system that quantifies the percentage of text in a submitted document that matches content in its database of academic papers, websites, and publications. It appears as a percentage-typically between 0% and 100%-and indicates how much of the work overlaps with existing sources.
The system scans submissions against a vast repository of over 160 million student papers, 100 billion web pages, and 100,000+ journals and publications. When a match is found, Turnitin flags the overlapping text and assigns a similarity score based on the proportion of matched content relative to the total word count.
A similarity score above 15% may trigger academic review, though context matters. A high score doesn't automatically mean plagiarism-it could reflect proper citation, common phrases, or standard terminology in a field. Educators assess both the score and the nature of matches before drawing conclusions.
When a student submits a paper through a Turnitin-integrated platform like Canvas, the software performs a real-time analysis. It breaks the document into segments, compares each segment against its database using algorithms that detect near-identical phrasing, and identifies sources with matching patterns.
Matches are grouped by source type: published articles, websites, student papers, or institutional repositories. Each match is highlighted in the report with a color-coded indicator and a link to the original source. The system also detects paraphrased content if the wording closely mirrors a source, though it's less precise with reworded passages.
For example, a 1,200-word research paper on climate change submitted to a university course returned a 28% similarity score. Upon inspection, 12% came from two cited journal articles, 8% from a government environmental report, and 8% from common scientific terminology like "carbon sequestration" and "global warming potential." Only 1% involved unattributed copying. The instructor accepted the paper after verifying citations were properly formatted.
Turnitin categorizes matches into distinct types. Understanding these helps distinguish between accidental overlap and intentional copying.
| Type | Description | Example |
|---|---|---|
| Published Material | Articles from journals, books, or magazines in Turnitin's database. | A quote from a Nature article on AI ethics. |
| Internet Sources | Web content from public sites, including Wikipedia and news outlets. | Paraphrased summary of a CNN piece on election fraud. |
| Student Papers | Previously submitted work from other students at the same or different institutions. | Same assignment from last semester's class. |
| Other Sources | Unidentified or low-visibility matches, often from obscure forums or unpublished drafts. | Reused paragraph from a personal blog. |
Not all matches indicate misconduct. Common phrases like "" or technical terms such as "quantitative analysis" appear frequently across disciplines and contribute to the score without violating academic norms. The presence of a match does not equate to plagiarism.
According to the University of California, Davis Writing Center, "A high similarity score is not synonymous with cheating. Instructors must evaluate the context, intent, and attribution of matched content."
Many students assume a score above 20% means they've plagiarized. This is incorrect. Turnitin does not flag individual sentences as "plagiarism"-it only highlights potential matches. The interpretation is left to the educator.
Another myth is that Turnitin detects all forms of plagiarism. It cannot identify ideas taken from memory or concepts derived from reading without direct quotation. It only detects textual reuse. A student who reads five sources and writes a synthesis in their own words may still show a 10% similarity due to shared terminology.
Also, Turnitin does not check for grammar, structure, or argument quality. A paper with perfect syntax and original analysis can have a high similarity score if it uses standard definitions or widely accepted facts. For example, defining "photosynthesis" as "the process by which green plants convert sunlight into energy" appears in thousands of biology textbooks and will consistently generate matches.
I once reviewed a 30-page literature review for a graduate course. The similarity score was 34%, but upon close inspection, 22% came from quoted material properly cited in APA format. The remaining 12% included standard phrases like "further research is needed" and "this study builds on prior work." The professor accepted the paper after confirming all sources were attributed.
Students should treat Turnitin similarity as a diagnostic tool, not a grading rubric. Submitting drafts early allows time to revise overlapping sections, especially when quoting or summarizing sources. Always cite sources, even when paraphrasing, and use quotation marks around direct text.
Instructors should avoid setting arbitrary thresholds like "no more than 10%." Instead, they should examine the report's breakdown and consider the discipline. A 25% score in a humanities paper may be acceptable if most matches are from cited sources. In STEM fields, higher similarity is common due to standardized terminology.
Some institutions use Turnitin's Originality Report to guide feedback. For instance, the University of Michigan's Writing Center recommends instructors use the report to prompt students to improve citation practices rather than penalize high scores.
Turnitin similarity is not a verdict-it's a signal. It supports academic integrity by helping educators identify potential issues, but it requires human judgment to interpret correctly. Context, citation accuracy, and disciplinary norms determine whether a match constitutes a violation.