Plagiarism checker

Yes, and then it does something different: it looks for sentences that repeat earlier sentences inside your own draft — a paragraph pasted twice, a reworded version of your own first page, padding. The result then says in as many words that nothing outside the draft was compared, because an empty source list is not a finding of originality.

No, and it says so rather than pretending. Searching the web needs a server, a crawl index and money; an offline page physically cannot. What it does is compare your draft against the text you paste — a chapter, an article, a previous submission, your own notes. That catches the copying students are actually caught for: from the material in front of them.

Each sentence is reduced to overlapping five-word shingles, and a draft sentence is compared with every source sentence using Jaccard similarity — the size of the overlap divided by the size of the union. Over the threshold (35% by default) it counts as a match and the pair is listed. Shingle matching survives small edits like swapped words or a changed tense.

Because a copier edits. Comparing exact strings finds only copy-paste, while a shingle overlap still sees the sentence when three words of fifteen changed. Lower the threshold in the dropdown to be stricter, but expect more false positives on common phrasing.

Yes — under six words they are skipped. “Yes, that is true.” overlaps with half the internet and would wreck the percentage. The report states how many sentences were comparable, so a small number is visible instead of hidden.

Two different things, both shown: how many comparable sentences overlapped a source, and how much of the draft’s word count sits inside those sentences. A 40% sentence match on a draft with three long sentences is not the same as 40% of the words — read both numbers before drawing a conclusion.