Plagiarism Side-by-Side Scanner

Paste two texts and detect matching n-grams (3-word sequences). Get a Jaccard similarity score and a list of overlapping phrases — all in your browser.

—

—

—

Similarity Breakdown

Matching Phrases

How This Comparison Works

Plagiarism detection is fundamentally about detecting shared n-grams — sequences of n consecutive words. If two texts share many 3-word sequences that are not common English phrases (like "in the end" or "on the other hand"), there is likely copying or close paraphrase.

The Algorithm

  1. Both texts are lowercased and stripped of punctuation.
  2. Words are tokenised and split into overlapping 3-grams.
  3. Common stop-word n-grams are filtered out — "of the", "in the", "to the" don't count as plagiarism.
  4. The Jaccard similarity is computed: |A ∩ B| ÷ |A ∪ B| — the ratio of shared n-grams to the total unique n-grams across both texts.
  5. Additionally, the fraction of B's n-grams that appear in A is computed — the "containment" score.

Interpreting the Scores

  • < 5% — Near-certainly independent. Shared n-grams are common phrases.
  • 5%–15% — Minor overlap. Could be coincidence, shared source material, or a single copied sentence.
  • 15%–35% — Significant overlap. Likely paraphrasing or partial copying.
  • > 35% — Heavy overlap. Likely substantial copying with light rewording.

Why "Plagiarism" Needs Human Judgment

Two texts can have 40% n-gram overlap without plagiarism: quotes from the same source, common technical definitions, or both describing the same API documentation. Two texts can have 8% overlap and still contain one flagrantly copied paragraph. Statistical overlap is a lead, not a verdict — always read the flagged phrases and judge them in context.

Paraphrasing Bypasses N-gram Detection

Adept paraphrasing — swapping synonyms, restructuring sentences, changing voice — can reduce n-gram overlap to near-zero while preserving the original meaning. This is called idea plagiarism, and no n-gram tool detects it. For that you need human editorial reading, not statistical analysis.

What This Tool Cannot Do

  • Compare against the entire web — that requires a search-index backend this tool deliberately does not have.
  • Detect cross-language plagiarism.
  • Detect paraphrased ideas (semantic plagiarism).
  • Detect AI-generated text (use the AI Burstiness Checker instead).

Plagiarism Scanner FAQ

Does this compare against the entire internet?

No. It compares two texts you paste in. Comparing against the web requires a search-index backend; this tool runs 100% locally for privacy.

What is n-gram overlap?

An n-gram is a sequence of n consecutive words. The tool splits both texts into 3-word sequences and measures how many are shared. High shared n-gram counts indicate copying or close paraphrase.

Can paraphrase fool this tool?

Yes — well-executed paraphrase reduces n-gram overlap to near zero while preserving meaning. This is "idea plagiarism" and requires human editorial judgment to detect.

What Jaccard score means plagiarism?

There is no universal threshold. In practice, 15%+ shared distinctive 3-grams is worth reviewing. 35%+ is strong evidence of copying.