A text similarity checker for comparing your own documents
This text similarity checker measures how much two to five texts overlap and shows you exactly where. Paste an essay and its source, two versions of an article, or a set of student answers, and it calculates a Jaccard similarity from shared word sequences (shingles), a cosine similarity from word counts, and how much of each text appears in the other. Every shared passage is highlighted side by side in matching colours, the longest runs of identical words are listed, and sentences that share most of their words are paired up.
Be clear about what it is: it compares only the texts you paste. It does not search the internet, books or academic databases, so it cannot tell you whether a text was copied from somewhere else you haven’t provided — that needs a full plagiarism service. It is ideal for checking a paraphrase against its source, spotting reused paragraphs between drafts or comparing submissions with each other, and it runs privately in your browser.
How to use the Text Similarity Checker
- Paste the first text into Text A and the second into Text B, or use the file button on a box to open a .txt, .md or .html file. Use Add text for up to five.
- Choose the word-sequence length (3 words is a good default) and the shortest passage you want highlighted.
- Leave “Ignore capital letters” on, and tick “Ignore common words” if you want to focus on content words.
- Read the Jaccard and cosine percentages, then look at the highlighted passages and the matching sentences.
- With three or more texts, click a cell in the similarity matrix to compare that pair in detail, then copy the report.
Text Similarity Checker features
Two complementary scores
Jaccard on word sequences catches copied phrasing; cosine on word counts catches shared vocabulary even when the order changes.
Direction-aware containment
“Of A found in B” shows how much of a short text sits inside a longer one, which Jaccard alone hides.
Side-by-side highlighting
Each shared passage gets its own colour in both texts; hover one to outline its twin.
Sentence matching
Pairs each sentence with its closest sentence in the other text and shows the overlap.
Up to five texts
A similarity matrix compares every pair at once — handy for checking a set of assignments against each other.
Open text files
Load .txt, Markdown or web pages (.html) into a box; the file is read on your device.
Private
Nothing is uploaded; close the tab and the texts are gone.
How the similarity scores are calculated
Both texts are split into words (punctuation is ignored). For Jaccard similarity, each text becomes the set of its overlapping n-word sequences, called shingles: “the quick brown fox jumps” has the 3-word shingles “the quick brown”, “quick brown fox” and “brown fox jumps”. Jaccard = shared shingles ÷ all distinct shingles in either text. Cosine similarity treats each text as a vector of word counts and measures the angle between them: 100% means the same words in the same proportions, 0% means no words in common.
Worked example: “the quick brown fox jumps over the lazy dog” and “the quick brown fox sleeps under the lazy dog” each have 7 trigrams; 3 are shared (“the quick brown”, “quick brown fox”, “the lazy dog”) out of 11 distinct, so Jaccard is 3/11 = 27.3%. Their word counts give a dot product of 9 and lengths of √11 each, so cosine is 9/11 = 81.8% — high, because almost the same words are used. Shared passages are found by greedy string tiling: the longest common runs of words are taken first, without overlaps.
How to read the results
Short texts swing a lot: one shared phrase in two sentences can mean 30%. Common stock phrases (“in conclusion”, “on the other hand”) create small overlaps between any two essays, which is why a percentage alone never proves copying.
| What you see | What it usually means |
|---|---|
| High Jaccard and long shared passages | Text copied with little change |
| Low Jaccard but high cosine | Same topic and vocabulary, reworded — a paraphrase or two texts about the same subject |
| High “of A found in B” but low Jaccard | A short text that sits largely inside a longer one |
| Matching sentences at 70–100% | Sentences lightly edited from one another |
| Everything near 0% | Unrelated texts |
Fair use of similarity scores
- Treat the percentages as a reason to look closer, not as a verdict. Quotations with citations and shared technical terms are legitimate overlap.
- When paraphrasing, aim for a low Jaccard at 3–5 words while keeping the meaning — and still cite the source. The paraphrasing tool can help you reword, and the citation generator formats references.
- To see every inserted and deleted word between two drafts, use the diff checker; for word statistics, the word frequency counter.
Text Similarity Checker FAQ
Is this a plagiarism checker?
Not a full one. The text similarity checker compares only the texts you paste with each other; it does not search the web, books or journals. Use it to compare an essay with a known source, drafts with each other, or several submissions.
Is this a free plagiarism checker that also detects AI writing?
It is free, but it does neither web searching nor AI detection. It compares the texts you give it and shows the copied passages between them. No tool can reliably prove that a text was written by AI.
What is a good similarity percentage?
There is no universal threshold. Unrelated texts on different topics score near 0%; two texts on the same subject often share 5–15% of 3-word sequences through common phrases; long highlighted passages matter more than the headline number.
What is the difference between Jaccard and cosine similarity?
Jaccard compares sets of word sequences, so it rewards identical phrasing. Cosine compares word frequencies regardless of order, so reworded text on the same topic still scores high.
Which n-gram size should I use?
Three to five words works well for detecting copied phrasing. One word measures shared vocabulary; longer sequences only match near-verbatim text.
Are my texts uploaded?
No. All comparison happens in your browser, and nothing is stored after you close the page.
Can I compare more than two texts?
Yes, up to five. A matrix shows the Jaccard and cosine scores for every pair; click a cell to see that pair’s highlighted passages and matching sentences.