An embedding turns content into a point in a high-dimensional space where distance means similarity. Embeddings power semantic search, recommendation, re-identification, and retrieval for RAG; you compare two items by measuring how close their vectors are.