The frequency of plagiarism identified by text-matching software in scientific articles: a systematic review and meta-analysis

The aim of this systematic review and meta-analysis is to determine the frequency of plagiarism in scientific papers estimated from publications that use text-matching software to identify plagiarism. For this purpose, a literature search of 39 bibliographic databases has been conducted and a total of 10,005 articles have been identified. Ten articles met the criteria for inclusion in the meta-analysis and they checked for plagiarism in 6459 already published articles or manuscripts submitted to journals or conferences. All articles assessed plagiarism in a two-step process, first identifying textual similarity based on text-matching software and second, additionally inspecting detected similarity in the human verification process. The result revealed that 18% (95% CI: 12-25%) of articles have instances of plagiarism. Subgroup analyses were conducted to explain the large variance in the results. Following factors were tested: the number of plagiarism criteria implemented during the human verification process, sample size, the country where the study was conducted, the scientific discipline of analyzed papers, and publication status of analyzed papers. Plagiarism rates were higher across studies with a smaller sample size (N < 500) or a larger number of plagiarism criteria used to identify plagiarism (4 or 5 criteria). In conclusion, text-matching software is effective in providing evidence for plagiarism; however, this includes only textually based cases of plagiarism, and the reliability of software results depends on additional human verification.

Automated summary

This systematic review and meta-analysis aimed to determine the frequency of plagiarism in scientific papers using text-matching software. The results showed that 18% of articles had plagiarism instances, and while the software was effective in providing evidence, additional human verification was crucial for reliability.