Source overlap
Paste several reports about the same event. This estimates which ones are restating one another rather than adding independent information. Runs in your browser; nothing is uploaded.
Loading model…
How it works: each snippet is turned into a vector with a small sentence-embedding model (all-MiniLM-L6-v2), and every pair is scored by cosine similarity. Above 0.85 is likely the same claim reworded; 0.70–0.85 is worth checking. It measures wording, not provenance — two genuinely independent reports of the same fact will also score high. Treat a flag as a prompt to trace the citations, not as proof.