Search indexes
Search indexes retrieve ranked matches and return a score that indicates how closely each result matches a query.
Unlike performance indexes, search indexes are queried explicitly rather than selected automatically by the Cypher® planner.
The Cypher SEARCH clause supports vector indexes from Neo4j 2026.01 and full-text indexes from Neo4j 2026.09.
Built-in procedures remain available for compatibility with earlier Neo4j versions and Cypher 5.
Two search indexes are available in Neo4j:
-
Full-text indexes provide lexical search over the content of
STRINGproperties and return relevance scores. -
Vector indexes provide approximate nearest-neighbor similarity search over vectors. Vector search is commonly used for semantic search when the indexed vectors encode semantic meaning.
Full-text and vector indexes can also be used together for hybrid search. This is useful when an application needs both lexical matches, such as exact terms, names, acronyms, or identifiers, and semantic matches from vector search over embeddings. When combining results from different search indexes, rank each source independently rather than comparing raw score values directly. For a worked example, see Developer Guide → Hybrid search.