What is semantic search?

Semantic search is search that finds results by meaning rather than by matching exact words. A search for “money back on a broken item” can find a page titled “refunds for damaged goods”, even though they share almost no words. It works by turning both the search and the stored content into embeddings, which are lists of numbers that represent meaning, and finding the closest matches.

Why it matters for a business

Semantic search is the retrieval step in most RAG systems, so it decides which passages a model answers from. It is good at paraphrase and weak at exact identifiers such as order numbers, product codes and names, which ordinary keyword search handles well, so production systems often combine the two. When it returns something similar-looking but wrong, such as the consumer returns policy for a wholesale question, the model answers fluently from the wrong page.

Go deeper

How AI uses tools and memory

Related terms

• Embeddings
• Vector database
• RAG (retrieval-augmented generation)
• Grounding

All terms, A to Z