Free Text Similarity Checker
Compare two texts in your browser with Jaccard and containment scores, shared passages highlighted in both and a sentence match table. No web search.
Loading tool…
How to measure wording overlap between two texts
Paste two texts to see how many word sequences they share, where the longest shared passages are and which sentences match, all calculated on your device.
- Paste the first text into Text A and the second into Text B, up to 30,000 characters each.
- Choose a shingle size of 3, 4 or 5 words. Smaller sizes catch shorter overlaps; larger sizes are stricter.
- Choose whether to ignore case, punctuation and common words such as the, and or of.
- Select Compare texts to see Jaccard similarity, containment in each direction and the number of shared shingles.
- Review the numbered passages highlighted in both texts and the sentence table, then copy or download the report.
When to compare two texts
Teachers comparing two essays, editors checking whether a guest post reuses an earlier article, and site owners looking for duplicate product descriptions all need the same answer: how much wording do these two texts have in common, and where is it?
The checker is also useful for your own work. Compare a rewrite with its source to confirm you changed enough, or compare two versions of a policy to see which paragraphs survived. For a line-by-line view of edits, the Text Diff tool is a better fit.
It compares only what you paste. It does not search the internet, so it cannot tell you whether either text was copied from somewhere else.
Worked example with the built-in texts
Select Load example to paste two short health paragraphs. With the defaults of 3-word shingles, ignore case and ignore punctuation on, Text A has 33 words and 31 distinct shingles, Text B has 35 words and 33 shingles, and 18 shingles are shared. Jaccard similarity is 18 divided by 46, or 39.1 percent. Text A found in Text B is 18 of 31, or 58.1 percent, and Text B found in Text A is 18 of 33, or 54.5 percent.
Two passages are highlighted: a 12-word run, Walking for thirty minutes a day is a simple way to start, and a 10-word run beginning Regular exercise improves heart health. The sentence table pairs A1 with B2 and A2 with B1 at 100 percent, because every word sequence in each Text A sentence appears inside the longer Text B sentence.
Switching to 5-word shingles leaves 14 shared of 29 and 31, so Jaccard falls to 30.4 percent while the same two passages remain.
How the scores are calculated
Each text is split into words, optionally lowercased with punctuation and common words removed. A shingle is every run of n consecutive words, so the sentence the cat sat down has two 3-word shingles. Jaccard similarity is the number of shared shingles divided by the number of distinct shingles in both texts combined. Containment divides the shared count by one text's own shingles.
Common passages are found by extending every shared shingle as far as the words keep matching, then keeping the longest runs that do not overlap. Sentence matches report, for each Text A sentence, the Text B sentence containing the largest share of its 3-word sequences, listed when that share is at least 50 percent.
Common mistakes to avoid
Do not read a low score as proof of originality. Paraphrased text, translated text or reordered ideas share few exact word sequences and will score low even when the content is borrowed.
Do not compare very short texts with large shingles. Two 20-word texts produce only a handful of 5-word shingles, so one change can swing the percentage sharply.
Remember that options change the numbers. Ignoring common words usually lowers overlap from stock phrasing, while keeping punctuation makes matches stricter. Note the settings when you share a result.
Useful tips and limitations
- Look at containment as well as Jaccard; a short quote copied into a long article shows high containment but a low Jaccard score.
- Turn on Ignore common words when two texts share stock phrasing such as of the and in the that inflates the overlap.
- Use 3-word shingles to catch short reused phrases and 5-word shingles to focus on clearly copied passages.
- Download the report to keep the scores, settings, passages and sentence matches together.
Common questions
Does the Text Similarity Checker search the internet for copied content?
No. It compares only the two texts you paste. It does not search the web, a database or previous submissions, so it cannot find sources you do not provide.
Are my texts uploaded when I use the Text Similarity Checker?
No. Tokenizing, shingling and matching all run in your browser. Nothing is sent to a server, and the texts are not saved after you leave the page.
Can the Text Similarity Checker detect paraphrased passages?
Only partly. It finds shared word sequences, so text rewritten with different words or a different order scores low even if the meaning is the same. Sentence matches can still surface lightly edited sentences.
What is the difference between Jaccard and containment in the Text Similarity Checker?
Jaccard compares shared shingles with all distinct shingles in both texts, so it is symmetric. Containment shows how much of one text appears in the other, which is better when the texts differ greatly in length.
Which shingle size should I pick in the Text Similarity Checker?
Use 3 words for short texts or to catch brief shared phrases. Use 4 or 5 words to ignore common short phrases and focus on longer copied passages.
How long can each text be in the Text Similarity Checker?
Each box accepts up to 30,000 characters. The comparison runs on your device, so very long texts may take a moment on older phones.
Your input is processed in your browser without a ToolWorld Online account. Copy or download anything you want to keep before leaving the page.