diff --git a/backend/alembic/versions/0009_search_suggestions.py b/backend/alembic/versions/0009_search_suggestions.py
index 4433ce9..33e9822 100644
--- a/backend/alembic/versions/0009_search_suggestions.py
+++ b/backend/alembic/versions/0009_search_suggestions.py
@@ -218,23 +218,19 @@ def upgrade() -> None:
                     FROM generate_series(1, length(word)) AS n
                 )
             FROM ts_stat(
-                'simple',
-                (
-                    SELECT to_tsvector(
-                        'simple',
-                        coalesce(title_zh, '') || ' ' ||
-                        coalesce(body_zh_text, '') || ' ' ||
-                        coalesce(commentary, '') || ' ' ||
-                        coalesce(commentary_meituan, '')
-                    )
-                    FROM articles
-                    WHERE title_zh IS NOT NULL
-                       OR body_zh_text IS NOT NULL
-                       OR commentary IS NOT NULL
-                       OR commentary_meituan IS NOT NULL
+                $$SELECT to_tsvector('simple',
+                    coalesce(title_zh, '') || ' ' ||
+                    coalesce(body_zh_text, '') || ' ' ||
+                    coalesce(commentary, '') || ' ' ||
+                    coalesce(commentary_meituan, '')
                 )
-            )
-            WHERE length(word) >= 2;  -- 过滤单字噪音(中文标点/单字停用词)
+                FROM articles
+                WHERE title_zh IS NOT NULL
+                   OR body_zh_text IS NOT NULL
+                   OR commentary IS NOT NULL
+                   OR commentary_meituan IS NOT NULL$$
+            ) AS s
+            WHERE length(s.word) >= 2;  -- 过滤单字噪音(中文标点/单字停用词)
 
         END;
         $$ LANGUAGE plpgsql;
diff --git a/backend/alembic/versions/0010_zhparser_chinese.py b/backend/alembic/versions/0010_zhparser_chinese.py
index 83c1b88..221e4fe 100644
--- a/backend/alembic/versions/0010_zhparser_chinese.py
+++ b/backend/alembic/versions/0010_zhparser_chinese.py
@@ -80,7 +80,8 @@ def upgrade() -> None:
             TRUNCATE search_keywords;
 
             -- ts_stat(query text) 接受 SQL 字符串,内部执行并聚合词频
-            -- 'a' = 任意权重(A/B/C/D 四档,这里聚合所有)
+            -- 单参(等价 mask='abcd',聚合所有权重)
+            -- ⚠️ 不能传 'a' mask — zhparser 不标 A 权重,会 0 行
             INSERT INTO search_keywords (keyword, source, weight, prefix_keys)
             SELECT
                 word,
@@ -91,8 +92,7 @@ def upgrade() -> None:
                     FROM generate_series(1, length(word)) AS n
                 )
             FROM ts_stat(
-                $q$
-                SELECT to_tsvector('chinese_zh',
+                $$SELECT to_tsvector('chinese_zh',
                     coalesce(title_zh, '') || ' ' ||
                     coalesce(body_zh_text, '') || ' ' ||
                     coalesce(commentary, '') || ' ' ||
@@ -102,10 +102,9 @@ def upgrade() -> None:
                 WHERE title_zh IS NOT NULL
                    OR body_zh_text IS NOT NULL
                    OR commentary IS NOT NULL
-                   OR commentary_meituan IS NOT NULL
-                $q$, 'a'
-            )
-            WHERE length(word) >= 2;
+                   OR commentary_meituan IS NOT NULL$$
+            ) AS s
+            WHERE length(s.word) >= 2;
         END;
         $func$ LANGUAGE plpgsql;
         """
@@ -135,8 +134,7 @@ def downgrade() -> None:
                     FROM generate_series(1, length(word)) AS n
                 )
             FROM ts_stat(
-                $q$
-                SELECT to_tsvector('simple',
+                $$SELECT to_tsvector('simple',
                     coalesce(title_zh, '') || ' ' ||
                     coalesce(body_zh_text, '') || ' ' ||
                     coalesce(commentary, '') || ' ' ||
@@ -146,10 +144,9 @@ def downgrade() -> None:
                 WHERE title_zh IS NOT NULL
                    OR body_zh_text IS NOT NULL
                    OR commentary IS NOT NULL
-                   OR commentary_meituan IS NOT NULL
-                $q$, 'a'
-            )
-            WHERE length(word) >= 2;
+                   OR commentary_meituan IS NOT NULL$$
+            ) AS s
+            WHERE length(s.word) >= 2;
         END;
         $func$ LANGUAGE plpgsql;
         """
diff --git a/backend/app/services/search.py b/backend/app/services/search.py
index 427e7cf..45eb43a 100644
--- a/backend/app/services/search.py
+++ b/backend/app/services/search.py
@@ -130,8 +130,9 @@ class SearchService:
     async def _fallback_keywords(self, q: str, limit: int) -> list[dict]:
         """回退:ts_stat 实时聚合(慢但能用)。
 
-        - 从 articles.title_zh + body_zh_text 实时 to_tsvector
-        - 适用:search_keywords 表空 + ts_stat 之前的全量聚合
+        - 从 articles.title_zh + body_zh_text 实时 to_tsvector(chinese_zh)
+        - 适用:search_keywords 表空 + worker 没刷新过
+        - ts_stat(text) 单参 — 第二参 weights mask 不能传 'a'(zhparser 不标 A 权重会 0 行)
         """
         from sqlalchemy import text
 
@@ -139,16 +140,13 @@ class SearchService:
             """
             SELECT word, nentry::int AS weight
             FROM ts_stat(
-                'simple',
-                (
-                    SELECT to_tsvector(
-                        'simple',
-                        coalesce(title_zh, '') || ' ' || coalesce(body_zh_text, '')
-                    )
-                    FROM articles
-                    WHERE title_zh IS NOT NULL OR body_zh_text IS NOT NULL
+                $$SELECT to_tsvector('chinese_zh',
+                    coalesce(title_zh, '') || ' ' || coalesce(body_zh_text, '')
                 )
-            )
+                FROM articles
+                WHERE title_zh IS NOT NULL OR body_zh_text IS NOT NULL
+                LIMIT 500$$
+            ) AS s
             WHERE word LIKE :prefix
             ORDER BY nentry DESC
             LIMIT :lim