How is the similarity score calculated?

It combines three measures: cosine similarity of word frequencies (30%), Jaccard similarity of the two vocabularies (20%) and the share of each text that sits inside a shared passage (50%, averaged over both texts). The individual numbers are shown next to the score so you can see what drives it.

Does it detect paraphrasing or AI rewrites?

Only partly. The tool measures word overlap, not meaning. Two texts that say the same thing in different words score low, and a heavily reworded copy may show as "some overlap". It is reliable for finding copied or lightly edited passages.

Is this a plagiarism checker?

It compares the two texts you paste and nothing else. It doesn't search the web or a database. Use it to check a draft against a specific source, compare two versions of a page, or find duplicated product descriptions.

What counts as a shared passage?

A run of at least the chosen number of consecutive words (5 by default) that appears in both texts, after applying the case, punctuation and stop word options. Each passage is extended as far as the texts keep matching, so you see the full copied section rather than overlapping fragments.

What does ignore stop words do?

It drops very common words such as "the", "and" and "of" before comparing. Cosine and Jaccard then reflect the meaningful vocabulary, and passages can match even when only small words were changed. It also makes passages easier to match by accident, so leave it off when you need strict copying evidence.

How long can the texts be?

Up to around 100,000 words each. Words and phrases are converted to numeric hashes, so the comparison stays fast. The highlighted view shows the first 150,000 characters of each text; the passage list and CSV export cover everything.

Are my texts uploaded?

No. Everything runs in your browser and nothing is stored.

Text Similarity Checker

Compare two texts and see how much wording they share. Get a similarity score, cosine and Jaccard metrics, and every shared passage highlighted in both texts.

100% free
No sign-up
Runs in your browser

Text A

376 characters · 69 words · 3 lines

Text B

349 characters · 63 words · 3 lines

Ignore caseIgnore punctuationIgnore stop words

Shortest shared passage

In words, 2 to 50.

Similarity

63%

Similar

Cosine

86%

Word frequency

Jaccard

58%

38 shared words

Passages

5+ words in a row

Text A found in text B49%

Share of its 69 compared words inside a shared passage.

Text B found in text A54%

Share of its 63 compared words inside a shared passage.

Phrase length	Shared	Of text A	Of text B
3 words	29	43%	48%
4 words	25	38%	42%
5 words	22	34%	37%
6 words	19	30%	33%
7 words	16	25%	28%
8 words	13	21%	23%

Phrases counted once each. This measures shared wording, not shared meaning, so a heavily paraphrased text scores low even if it says the same thing.

Shared passages in both texts

Matching numbers mark the same passage. Tap one to pick it out.

Text A

Instagram DMs are where most sales conversations start. ¹Someone sees a post, comments a keyword and waits for a reply. If the reply takes a day, the moment has passed and ²the buyer has moved on to the next account in their feed. Automating the first reply fixes that. ³The customer gets the link or price straight away, and you only step in when a real conversation needs you.

Text B

Most sales conversations now start in Instagram DMs. ¹Someone sees a post, comments a keyword and waits for a reply. When the answer takes hours, ²the buyer has moved on to the next account in their feed. An automated first reply solves this. ³The customer gets the link or price straight away, and your team joins only when a person is really needed.

Longest shared passages

What the similarity score measures

The checker compares the words two texts use and the order they use them in. It doesn't use AI and it doesn't understand meaning: "the price went up" and "it costs more now" share almost no words, so they score as different. What it does well is find copied, reused and lightly edited text, quickly and precisely.

The four measures

Cosine similarity compares how often each word appears in both texts. Texts on the same topic with similar vocabulary score high, even if the sentences differ.
Jaccard similarity is the number of distinct words both texts use, divided by the number of distinct words in either.
Shared n-grams count the 3- to 8-word phrases found in both texts. Long shared phrases are a strong sign of copying, because different writers rarely produce the same eight words in a row.
Shared passages are the longest runs of matching words, at least the length you choose. The coverage bars show how much of each text sits inside them.

The overall score is 30% cosine, 20% Jaccard and 50% passage coverage. Below 20% the texts are very different, 20-44% share some wording, 45-74% are similar and 75% or more are near duplicates.

Good uses

Check a guest post, product description or caption against the source it was based on.
Find duplicate content across two pages of your site before it causes SEO problems.
Compare two versions of a script, DM template or contract to see how much was reused.

How it works

How to use it

Step 1
Paste two texts
Add the original on the left and the text you want to compare on the right. Results update as you type.
Step 2
Set the options
Ignore case, punctuation or common stop words, and choose how many words in a row count as a shared passage.
Step 3
Read the overlap
See the overall score, the individual metrics and every shared passage highlighted in both texts with matching numbers.

FAQs

Frequently asked questions

Answers to common questions about the Text Similarity Checker.

: It combines three measures: cosine similarity of word frequencies (30%), Jaccard similarity of the two vocabularies (20%) and the share of each text that sits inside a shared passage (50%, averaged over both texts). The individual numbers are shown next to the score so you can see what drives it.
: Only partly. The tool measures word overlap, not meaning. Two texts that say the same thing in different words score low, and a heavily reworded copy may show as "some overlap". It is reliable for finding copied or lightly edited passages.
: It compares the two texts you paste and nothing else. It doesn't search the web or a database. Use it to check a draft against a specific source, compare two versions of a page, or find duplicated product descriptions.
: A run of at least the chosen number of consecutive words (5 by default) that appears in both texts, after applying the case, punctuation and stop word options. Each passage is extended as far as the texts keep matching, so you see the full copied section rather than overlapping fragments.
: It drops very common words such as "the", "and" and "of" before comparing. Cosine and Jaccard then reflect the meaningful vocabulary, and passages can match even when only small words were changed. It also makes passages easier to match by accident, so leave it off when you need strict copying evidence.
: Up to around 100,000 words each. Words and phrases are converted to numeric hashes, so the comparison stays fast. The highlighted view shows the first 150,000 characters of each text; the passage list and CSV export cover everything.
: No. Everything runs in your browser and nothing is stored.

More free tools

Try another tool

View all free tools →

Reply to every DM without copy-pasting

Hypello sends comment-to-DM replies and keyword-triggered DMs automatically, and its AI agents collect leads and offer discount codes in the conversation. Start free with 1,000 DMs a month.

Get Started For Free View Pricing →