Semantic search matches queries to results based on meaning rather than keyword overlap. "How do I cancel my subscription?" surfaces documents about "account termination" and "membership cancellation" — even if the exact query words never appear.
How it works: both the query and the document corpus are converted to embeddings via a language model, stored in a vector database, and compared using cosine similarity or dot product. The top results are the ones whose meaning is closest to the query.
Semantic search dominates 2020s information retrieval — it powers internal knowledge base search, ecommerce discovery, legal precedent lookup, and every RAG system. Best practice combines semantic search with keyword search (hybrid search) for the strongest results.