diff --git a/cl/search/tests/tests.py b/cl/search/tests/tests.py
index 82784ce5ea..2ca99e9b9a 100644
--- a/cl/search/tests/tests.py
+++ b/cl/search/tests/tests.py
@@ -422,13 +422,18 @@ def setUpTestData(cls):
             sub_opinions=RelatedFactory(
                 OpinionWithChildrenFactory,
                 factory_related_name="cluster",
-                html_columbia="<p>Code, &#167; 1-815</p>",
+                html_columbia="<p>Code, &#167; 1-815 Lorem §247</p>",
             ),
             precedential_status=PRECEDENTIAL_STATUS.PUBLISHED,
         )
         OpinionClusterFactoryWithChildrenAndParents(
             case_name="Strickland v. Lorem.",
             docket=DocketFactory(court=cls.court, docket_number="123456"),
+            sub_opinions=RelatedFactory(
+                OpinionWithChildrenFactory,
+                factory_related_name="cluster",
+                plain_text="Random plain_text",
+            ),
             precedential_status=PRECEDENTIAL_STATUS.PUBLISHED,
         )
         OpinionClusterFactoryWithChildrenAndParents(
@@ -436,6 +441,11 @@ def setUpTestData(cls):
             docket=DocketFactory(
                 court=cls.child_court_1, docket_number="34-2535"
             ),
+            sub_opinions=RelatedFactory(
+                OpinionWithChildrenFactory,
+                factory_related_name="cluster",
+                plain_text="Lorem 247",
+            ),
             precedential_status=PRECEDENTIAL_STATUS.PUBLISHED,
         )
         OpinionClusterFactoryWithChildrenAndParents(
@@ -443,6 +453,11 @@ def setUpTestData(cls):
             docket=DocketFactory(
                 court=cls.child_court_2_2, docket_number="36-2000"
             ),
+            sub_opinions=RelatedFactory(
+                OpinionWithChildrenFactory,
+                factory_related_name="cluster",
+                plain_text="Random plain_text",
+            ),
             precedential_status=PRECEDENTIAL_STATUS.PUBLISHED,
         )
 
@@ -451,6 +466,11 @@ def setUpTestData(cls):
             docket=DocketFactory(
                 court=cls.child_gand_2, docket_number="38-1000"
             ),
+            sub_opinions=RelatedFactory(
+                OpinionWithChildrenFactory,
+                factory_related_name="cluster",
+                plain_text="Random plain_text",
+            ),
             precedential_status=PRECEDENTIAL_STATUS.PUBLISHED,
         )
         call_command(
@@ -838,6 +858,17 @@ def test_raise_forbidden_error_on_depth_pagination(self) -> None:
         )
         self.assertEqual(r.status_code, HTTPStatus.FORBIDDEN)
 
+    async def test_avoid_splitting_terms_on_phrase_queries(self) -> None:
+        """Can we avoid splitting words in phrase queries such as "§247"?"""
+
+        # A search for "Lorem §247" shouldn't match "Lorem 247"
+        r = await self.async_client.get(
+            reverse("show_results"), {"q": '"Lorem §247"'}
+        )
+        actual = self.get_article_count(r)
+        self.assertEqual(actual, 1)
+        self.assertIn("1:21-cv-1234", r.content.decode())
+
 
 class SearchAPIV4CommonTest(ESIndexTestCase, TestCase):
     """Common tests for the Search API V4 endpoints."""
diff --git a/cl/settings/third_party/elasticsearch.py b/cl/settings/third_party/elasticsearch.py
index 7a1ec6b779..656ada84a1 100644
--- a/cl/settings/third_party/elasticsearch.py
+++ b/cl/settings/third_party/elasticsearch.py
@@ -125,6 +125,7 @@
         "filter": {
             "custom_word_delimiter_filter": {
                 "type": "word_delimiter",
+                "type_table": ["§ => ALPHA"],
                 "split_on_numerics": False,
                 "preserve_original": True,
             },