What's the difference between edit-distance similarity and word-overlap (Jaccard) similarity?

Edit-distance similarity is character-level: it counts the minimum number of single-character insertions, deletions, or substitutions needed to turn one text into the other, then expresses that as a percentage of the longer text's length. It's sensitive to word order and small phrasing changes. Word-overlap (Jaccard) similarity ignores order entirely and just compares the set of words each text uses — so two sentences with the same words shuffled into a different order can score high on Jaccard but noticeably lower on edit distance. Use edit distance when phrasing and order matter (e.g. comparing document revisions); use word overlap when you care about shared vocabulary or topic regardless of structure.

What counts as "similar enough" for my use case?

There's no universal cutoff — it depends on what you're checking. For catching accidental duplicate content (e.g. two product descriptions or two form submissions), scores above roughly 90% usually mean near-duplicates. For comparing document revisions, a lower score is expected and just tells you how much changed. For a rough plagiarism screen, this tool can flag two pasted passages as suspiciously similar, but it is not a full plagiarism detector — it only compares the two texts you provide, it does not search the web or any external corpus, so a text with 0% similarity here could still be plagiarized from a source you haven't pasted in.

Is the comparison case-sensitive?

By default no — comparisons lowercase both texts first, so "Hello" and "hello" count as identical. Toggle "Case sensitive" above the text boxes to switch to an exact, case-sensitive comparison for both the edit-distance and word-overlap metrics.

Is there a limit on how much text I can compare?

Word-based (Jaccard) similarity has no practical limit. The character-level edit-distance calculation is a dynamic-programming algorithm that gets slower as texts grow, so above a combined 20,000 characters across both boxes, the tool skips the exact edit-distance pass and shows only the word-based similarity, with a note explaining why — this keeps the page responsive instead of freezing on very long inputs.

Is my text uploaded anywhere?

No. Both texts are compared entirely in your browser using JavaScript — nothing you paste is sent to a server or stored anywhere.

TEXT SIMILARITY CHECKER

Compare two texts and get an instant similarity score across edit-distance and word-overlap metrics. Nothing leaves your browser.

Text A

Text B

These texts are 88% similar — very similar.

Edit-distance similarity88.1%Character-level: how many single-character edits separate the two texts. Sensitive to word order and small phrasing changes.

Word overlap (Jaccard)76.5%Word-level: shared vocabulary regardless of order. Two texts with swapped word order can still score high here.

Shared words132 unique to A, 2 unique to B (15 vs 15 distinct words total).

How to check text similarity

1
Paste the first text into the Text A box and the second into the Text B box.
2
Optionally toggle "Case sensitive" if capitalization differences should count as a difference.
3
Read the similarity percentage and interpretation, then check the edit-distance, word-overlap, and shared-word breakdown for more detail.

Questions

What's the difference between edit-distance similarity and word-overlap (Jaccard) similarity?: Edit-distance similarity is character-level: it counts the minimum number of single-character insertions, deletions, or substitutions needed to turn one text into the other, then expresses that as a percentage of the longer text's length. It's sensitive to word order and small phrasing changes. Word-overlap (Jaccard) similarity ignores order entirely and just compares the set of words each text uses — so two sentences with the same words shuffled into a different order can score high on Jaccard but noticeably lower on edit distance. Use edit distance when phrasing and order matter (e.g. comparing document revisions); use word overlap when you care about shared vocabulary or topic regardless of structure.
What counts as "similar enough" for my use case?: There's no universal cutoff — it depends on what you're checking. For catching accidental duplicate content (e.g. two product descriptions or two form submissions), scores above roughly 90% usually mean near-duplicates. For comparing document revisions, a lower score is expected and just tells you how much changed. For a rough plagiarism screen, this tool can flag two pasted passages as suspiciously similar, but it is not a full plagiarism detector — it only compares the two texts you provide, it does not search the web or any external corpus, so a text with 0% similarity here could still be plagiarized from a source you haven't pasted in.
Is the comparison case-sensitive?: By default no — comparisons lowercase both texts first, so "Hello" and "hello" count as identical. Toggle "Case sensitive" above the text boxes to switch to an exact, case-sensitive comparison for both the edit-distance and word-overlap metrics.
Is there a limit on how much text I can compare?: Word-based (Jaccard) similarity has no practical limit. The character-level edit-distance calculation is a dynamic-programming algorithm that gets slower as texts grow, so above a combined 20,000 characters across both boxes, the tool skips the exact edit-distance pass and shows only the word-based similarity, with a note explaining why — this keeps the page responsive instead of freezing on very long inputs.
Is my text uploaded anywhere?: No. Both texts are compared entirely in your browser using JavaScript — nothing you paste is sent to a server or stored anywhere.