How much plagiarism is acceptable in a research paper?
Short answer: none — but that is not the same as a 0% similarity score. Plagiarism is using someone else's words or ideas without credit, and no amount of that is acceptable. A similarity score measures something different: how much of your text matches other documents, including properly quoted and cited passages, standard phrases and your own reference list. Journals do not publish one universal limit; editors read what matched. The few formal limits that exist — such as India's UGC levels for theses — come with exclusions for exactly those legitimate matches.
Similarity is not plagiarism
A checker reports overlap. Whether that overlap is plagiarism depends on what matched and whether it is credited:
- Acceptable: quotations in quotation marks with a citation, the reference list, names of methods, datasets and instruments, and common phrases of your field.
- Needs attention: a long run of text from one source, a paraphrase that follows the source sentence by sentence, or an idea with no citation.
- Self-overlap: text reused from your own earlier paper or thesis without saying so — journals treat this as a separate problem.
How editors read a similarity report
Many publishers check submissions through Crossref's Similarity Check service, which runs iThenticate. Editors look less at the total than at two things: whether a single source accounts for a large block of text, and whether the matched passages sit in the introduction, results or discussion — where your own words are expected — rather than in the methods or references. A paper at a moderate total spread across many tiny matches is usually fine; a paper with one lifted paragraph is not, even if its total looks low.
Where formal limits exist
Some institutions write thresholds into their rules. The best-known is India's UGC 2018 regulation for theses and dissertations, which sets levels of similarity and the penalties for each — see the UGC plagiarism rules guide for the levels and what the check excludes. Universities and journals that use a number usually apply it after excluding quotes, references and small matches, so compare like with like: your score with those exclusions, not the raw total.
Reading your own score
- Exclude the reference list and properly quoted text, as editors do.
- Sort the remaining matches by size and look at the largest single sources first.
- For each block, decide: cite it, quote it, or rewrite it in your own structure and cite it.
- Check for your own preprint or thesis — explain it to the editor rather than rewriting it.
- Re-check the revised version before you submit.
The plagiarism checker shows every match next to its source so you can make those calls, and the guide on reducing similarity the right way walks through each fix with an example.
What not to do
Swapping words for synonyms to push a number down does not remove plagiarism — the idea is still uncredited, and modern checkers compare meaning as well as wording. Hiding text from the checker with white characters, images of text or invisible symbols is treated as misconduct when it is found. Fix the credit, not the score.