Semantic Similarity Checker
Paste two texts. This compares them by meaning using an AI model that runs entirely in your browser — the first click downloads a small model (roughly 25-90MB depending on your browser's cache), then everything runs locally.
About this tool
Word-matching comparison tools miss paraphrased duplicates — text reworded with different vocabulary but the same underlying meaning. This tool instead converts each sentence into a numerical representation (an "embedding") using a small AI language model, then measures how close those representations are in meaning-space. Two sentences that mean roughly the same thing land close together even if they don't share a single word.
Why doesn't a completely unrelated sentence score 0%?
Sentence embeddings don't have a true zero — even very different sentences typically score somewhere in the 20-40% range because they share basic linguistic structure. Treat anything under about 45% as "not meaningfully related" rather than expecting it to approach 0%.
Is my text uploaded anywhere?
No. The AI model downloads once to your browser (and is cached for next time), and every comparison after that runs locally on your device. Nothing you paste is sent to a server.
Why did the first comparison take a while?
The first click downloads the AI model itself. Subsequent comparisons on the same visit — or on a later visit, since browsers cache it — are close to instant.