How vector search works
- 1Every document, or every chunk of a document, is turned into an embedding and stored in an index.
- 2The question is turned into an embedding the same way.
- 3The index returns the chunks closest to the question, measured by distance between the vectors.
- 4Those chunks are passed to the language model, which writes the answer.
Because closeness reflects meaning, a question about cheap flights can match a page that only says budget airfares.
Vector search vs keyword search
- Keyword search matches the words used. It is precise for names, codes and exact phrases.
- Vector search matches meaning. It handles synonyms, paraphrases and different languages better.
- Most AI systems combine both, which is called hybrid search.
What it means for brands
- Write each section so it is clearly about one thing; focused passages produce focused embeddings that match the right questions.
- Use the words your buyers use as well as your own product terms, so both keyword and vector matching work.
- Keep names, prices and specs in plain text near the claims they support.