Semantic search retrieves results based on meaning rather than literal keyword match, typically by computing similarity between vector embeddings of the query and indexed content.
Why it matters
Classic keyword search (Lucene, Elastic, BM25) excels at exact terms but fails when the same concept is phrased differently. Semantic search bridges that gap: a search for 'how to onboard new employees' surfaces a document titled 'employee orientation playbook' even with no shared keywords.
In practice, hybrid search (semantic + keyword) usually outperforms either alone. Keyword anchors precision; embeddings handle paraphrase. Tools like Pith, Glean, and modern enterprise search default to hybrid.
How Pith relates
Pith's search is hybrid by default: vector embeddings of bookmarks plus full-text indexing on titles and summaries. The wiki's RAG retrieval also uses semantic similarity to assemble each page's citation set.
See also
Last reviewed: 10 May 2026 · Licensed CC BY 4.0 · cite freely with attribution to Pith.