Web-wide plagiarism detection requires a server and a search index of billions of pages, which a client-side tool cannot have. Instead, this tool does something it can do honestly and privately: it compares your main text against the source material you provide — one source or many — and finds the passages they share. It is ideal for checking a draft against the papers you cited, comparing two documents, or catching repetition within one document.

Each text is normalised (lower-cased, punctuation stripped) and split into overlapping k-word shingles. The tool finds shingles your main text shares with each source, merges contiguous matches into passages, and measures coverage — the percentage of your words inside a matched passage. The headline score is coverage-led, nudged by the strongest Jaccard set-similarity so near-identical documents still read high. A per-source breakdown shows each source’s coverage and Jaccard separately.

Every comparison is computed locally in JavaScript — large inputs run in a Web Worker so the editor stays smooth. Nothing is sent to a server, there is no account and no tracking, and the page works offline after first load, so you can safely check confidential drafts, unpublished research or student work.

Sensitivity controls how many consecutive words must match before a passage counts. A small shingle size (for example 4 words) catches short reused phrases but may flag common idioms; a larger size (7–8 words) reports only longer, more deliberate overlaps. Tune it to balance recall against false positives for your text.