Text Similarity Checker

This tool breaks both texts into overlapping three-word sequences called trigrams, builds a frequency vector for each text, and calculates the cosine of the angle between the two vectors. A score of 0% means no trigrams are shared; 100% means every trigram appears in both texts at the same frequency. It also scans for runs of five or more consecutive matching words and highlights them as matching passages.

There is no hard threshold, but scores above 50% suggest enough shared phrasing to risk a duplicate-content flag from search engines. Scores above 70% usually mean one text is closely derived from the other. Below 30% is generally considered unique. Two articles on the same topic will naturally share some vocabulary, but shared three-word phrases indicate structural similarity, which is what search engines look for.

Cosine similarity measures the angle between two numerical vectors. In this tool each text is converted into a vector of trigram frequencies. If the vectors point in the same direction the cosine is 1 (identical); if they are perpendicular the cosine is 0 (no overlap). The percentage shown is the cosine value multiplied by 100. It is widely used in information retrieval and natural-language processing because it focuses on the pattern of term usage rather than document length.

This tool compares two texts you paste in. It does not search the web or any database. For full plagiarism detection you need a service that indexes billions of pages. This tool is designed for a different use case: comparing two known texts -- for example, checking whether a rewrite is different enough from the original, or verifying that two product descriptions on your own site are not too similar.

A high score means the two texts share a large proportion of their three-word phrases. If the texts are on the same site, this can cause search engines to treat one as a duplicate and suppress it from results. Consider rewriting shared passages, merging the two pages into one, or adding a canonical tag to tell search engines which version to index.