What it means
Semantic search finds content based on meaning rather than exact word matches, typically by comparing embeddings. It can match a question to a relevant document even when they use different wording.
Real-world example
In an archive's online catalogue, a user searches for "women's voting rights". Semantic search also returns records described as "Frauenwahlrecht", "suffrage movement" and "1918 franchise reform", which a word-matching search would have missed.
Why it matters for AI adoption
Semantic search makes knowledge more findable for people who do not know an organisation's internal vocabulary. But it can return results that are related rather than precise, and it can struggle with exact identifiers like file numbers or legal references. Many good systems combine semantic and keyword search.
Often confused with
Keyword search matches the exact words (or word forms) in a query. Semantic search matches meaning. Keyword search is better for names, codes and exact phrases; semantic search is better for natural-language questions. Combining both is called hybrid search.
Questions to ask
- Does the system combine semantic and keyword search, and how are results ranked?
- How do we test whether search results are relevant for our users' real questions?