Semantic search aims to return results that are conceptually relevant, even when the query and the target text use different wording. It is especially useful when users express the same idea in many different ways.
That usually means representing queries and documents with embeddings, then comparing them with a similarity metric such as cosine similarity.