diff --git a/main.go b/main.go
index 32033ff..0f155e8 100644
--- a/main.go
+++ b/main.go
@@ -3556,13 +3556,18 @@ func (se *SearchEngine) searchAllTables(tokens []string, limit int) ([]SearchRes
 	slog.Debug("Built FTS query", "fts_query", ftsQuery)
 
 	// Use pure FTS5 search with bm25() column weights for title prioritization
-	// bm25(search, 1.0, 3.0, 1.0, 1.0, 1.0, 1.0) weights: type, title(3x), body, url, repository, author
-	// Multiply by boost to prioritize user's authored content (2x boost)
+	// bm25(search, 1.0, 5.0, 1.0, 1.0, 1.0, 1.0) weights: type, title(5x), body, url, repository, author
+	// Multiply by boost for user's contributed repos (2x), state_boost for open items (1.5x),
+	// and recency_boost based on created_at age (<30d: 1.0, 30-180d: 0.85, >180d: 0.7)
 	query := `
 		SELECT type, title, body, url, repository, author, created_at, state
 		FROM search 
 		WHERE search MATCH ?
-		ORDER BY (bm25(search, 1.0, 3.0, 1.0, 1.0, 1.0, 1.0) * boost)
+		ORDER BY (bm25(search, 1.0, 5.0, 1.0, 1.0, 1.0, 1.0) * boost * 
+			CASE WHEN state = 'open' THEN 1.5 ELSE 1.0 END *
+			CASE WHEN julianday('now') - julianday(created_at) < 30 THEN 1.0 
+			     WHEN julianday('now') - julianday(created_at) < 180 THEN 0.85 
+			     ELSE 0.7 END)
 		LIMIT ?`
 
 	slog.Debug("Executing FTS query", "sql", query, "search_table", "search", "fts_query", ftsQuery, "limit", limit)
diff --git a/main.md b/main.md
index 9280435..85970f9 100644
--- a/main.md
+++ b/main.md
@@ -1201,7 +1201,7 @@ Where `<invalid_fields>` is a comma-separated list of invalid fields, and `<avai
 Next, prepare the FTS5 search query using the `search` table. Build the query with:
 
 - Use FTS5 MATCH operator for the search query
-- Order by `bm25(search)` for optimal relevance ranking (titles are weighted 3x higher)
+- Order by `bm25(search)` for optimal relevance ranking (titles are weighted 5x higher in BM25 base scoring, then additional multipliers applied for state and recency)
 - Limit to 20 results
 - Use the unified SearchEngine implementation shared with the UI
 
@@ -1434,11 +1434,18 @@ const SCHEMA_GUID = "550e8400-e29b-41d4-a716-446655440001" // Change this GUID o
 - FTS5 virtual table for full-text search across discussions, issues, and pull requests
 - Indexed columns: `type`, `title`, `body`, `url`, `repository`, `author`
 - Unindexed columns: `created_at`, `state`, `boost`
-- `boost`: Numeric value (e.g., `1.0`, `2.0`) used to multiply BM25 scores for ranking
-- Uses `bm25(search, 1.0, 2.0, 1.0, 1.0, 1.0, 1.0)` ranking with 2x title weight for relevance scoring
-- Search results should be ordered by: `(bm25(search) * boost)` for optimal relevance
-  - Items from user's repositories get 2x boost, ensuring they appear higher in results
-  - This approach is more flexible than boolean flags and allows for future ranking adjustments
+- `boost`: Numeric value (1.0 or 2.0) used to multiply BM25 scores for ranking, stored at index time (2.0 for user's contributed repos, 1.0 otherwise)
+- Uses `bm25(search, 1.0, 5.0, 1.0, 1.0, 1.0, 1.0)` ranking with 5x title weight for relevance scoring
+- Search results should be ordered by: `(bm25(search) * boost * state_boost * recency_boost)` for optimal relevance
+- **Ranking factors:**
+  - **Title weight (5x):** Title matches are weighted 5x higher than other fields in BM25 scoring (column order: type=1.0, title=5.0, body=1.0, url=1.0, repository=1.0, author=1.0)
+  - **User-contributed repos (2x):** Items from repositories where the user has contributed get 2x boost (stored in `boost` column at index time)
+  - **Open state (1.5x):** Open items get 1.5x boost at query time: `CASE WHEN state = 'open' THEN 1.5 ELSE 1.0 END`
+  - **Recency decay:** Time-based decay calculated at query time based on `created_at`:
+    - Recent (<30 days): 1.0 (full score)
+    - Medium (30-180 days): 0.85
+    - Older (>180 days): 0.7
+    - SQL: `CASE WHEN julianday('now') - julianday(created_at) < 30 THEN 1.0 WHEN julianday('now') - julianday(created_at) < 180 THEN 0.85 ELSE 0.7 END`
 
 #### table:schema_version