Your agent runs two searches at once and merges the results. Keyword search catches exact terms like error codes and product names; semantic search catches meaning, so "why was my card rejected" finds an article titled "Failed payments".
The two ranked lists are fused, then weighted by how much each piece of content can be trusted — hand-written articles rank above scanned pages, which rank above AI-generated drafts.
Every answer cites the articles it came from, and when nothing confident is found the agent says so rather than guessing. See How your agent finds answers.