diff --git a/docs/OPENGRAPH.md b/docs/OPENGRAPH.md
index 1b65044..926395c 100644
--- a/docs/OPENGRAPH.md
+++ b/docs/OPENGRAPH.md
@@ -29,8 +29,9 @@ data. Mixed imports such as `SharpHound.zip MSSQLHound.zip` are treated as a
 BloodHound data import because they can establish or refresh the AD baseline.
 
 `metadata` may be omitted. If present, it must be a JSON object. When
-`metadata.source_kind` is present, it must be a non-empty string matching the
-same safe identifier rule as labels.
+`metadata.source_kind` is present, it must be a string. Empty or whitespace-only
+values are treated as omitted; non-empty values must match the same safe
+identifier rule as labels.
 
 If BloodHound CE analysis takes longer than the default 600 seconds, set
 `ADPF_BH_INGESTION_TIMEOUT_SECONDS` to a positive integer number of seconds.
@@ -140,8 +141,9 @@ Property endpoint example:
 ## Stubs
 
 Endpoint stubs use `metadata.source_kind` as their label when it is present. If
-`source_kind` is missing, the importer uses `OpenGraph_Stub`. Invalid or empty
-`source_kind` values are rejected before any OpenGraph writes are attempted.
+`source_kind` is missing, empty, or whitespace-only, the importer uses
+`OpenGraph_Stub`. Invalid non-empty `source_kind` values are rejected before any
+OpenGraph writes are attempted.
 
 Stubs are created only for `match_by: "id"` endpoints that do not supply an
 endpoint `kind`. A typed endpoint that does not resolve is left unmatched rather
@@ -300,8 +302,9 @@ semantics. The upstream docs are:
 
 Intentional ADPathfinder differences:
 
-- `metadata.source_kind` may be omitted. If present, it must be a non-empty safe
-  identifier because ADPathfinder uses it as the fallback endpoint-stub label.
+- `metadata.source_kind` may be omitted. Empty or whitespace-only values are
+  treated as omitted. Non-empty values must be safe identifiers because
+  ADPathfinder uses them as fallback endpoint-stub labels.
 - ADPathfinder does not append `source_kind` to every imported node.
 - Missing or empty `node.kinds` falls back to `Base`; generic Base-only nodes
   are merged as `OpenGraph_Stub`.
diff --git a/modules/BloodhoundImporter.py b/modules/BloodhoundImporter.py
index 9420f48..e3eabce 100644
--- a/modules/BloodhoundImporter.py
+++ b/modules/BloodhoundImporter.py
@@ -526,17 +526,7 @@ def _validate_opengraph_payload(
         metadata = data.get('metadata', {})
         if not isinstance(metadata, dict):
             raise ValueError(f"OpenGraph {source_name}: metadata must be an object")
-        source_kind = metadata.get('source_kind', '')
-        if 'source_kind' in metadata and not isinstance(source_kind, str):
-            raise ValueError(
-                f"OpenGraph {source_name}: metadata.source_kind must be a string"
-            )
-        if 'source_kind' in metadata:
-            source_kind = self._safe_opengraph_identifier(
-                source_kind,
-                'source_kind',
-                f"OpenGraph {source_name}: metadata.source_kind",
-            )
+        source_kind = self._validated_source_kind(metadata, source_name)
 
         if merge_mode == 'opengraph' and source_kind in reserved_labels():
             raise ValueError(
@@ -604,6 +594,20 @@ def _validate_opengraph_payload(
                     "properties must be an object"
                 )
 
+    def _validated_source_kind(self, metadata: dict, source_name: str) -> str:
+        source_kind = metadata.get('source_kind', '')
+        if 'source_kind' in metadata and not isinstance(source_kind, str):
+            raise ValueError(
+                f"OpenGraph {source_name}: metadata.source_kind must be a string"
+            )
+        if not source_kind or not source_kind.strip():
+            return ''
+        return self._safe_opengraph_identifier(
+            source_kind,
+            'source_kind',
+            f"OpenGraph {source_name}: metadata.source_kind",
+        )
+
     def _validate_opengraph_endpoint(
         self,
         source_name: str,
@@ -688,9 +692,7 @@ def _stub_label_for_source_kind(self, data: dict) -> str:
         source_kind = metadata.get('source_kind', '')
         if 'source_kind' in metadata and not isinstance(source_kind, str):
             raise ValueError("OpenGraph metadata.source_kind must be a string")
-        if 'source_kind' in metadata and not source_kind:
-            raise ValueError("OpenGraph metadata.source_kind must be a non-empty string")
-        if source_kind:
+        if source_kind and source_kind.strip():
             label = safe_cypher_identifier(source_kind, 'source_kind')
             if label in reserved_labels():
                 raise ValueError(
diff --git a/tests/test_importer_opengraph.py b/tests/test_importer_opengraph.py
index 1a09a08..1a5d4da 100644
--- a/tests/test_importer_opengraph.py
+++ b/tests/test_importer_opengraph.py
@@ -1154,11 +1154,12 @@ def test_opengraph_rejects_present_non_string_source_kind_before_query(source_ki
     assert conn.queries == []
 
 
-def test_opengraph_rejects_empty_source_kind_before_query():
-    conn = RecordingConnection()
+@pytest.mark.parametrize("source_kind", ["", "   "])
+def test_opengraph_treats_blank_source_kind_as_omitted(source_kind):
+    conn = RecordingConnection(responses=[[{"c": 0}]])
     importer = BloodhoundImporter(conn)
     data = {
-        "metadata": {"source_kind": ""},
+        "metadata": {"source_kind": source_kind},
         "graph": {
             "nodes": [
                 {
@@ -1167,16 +1168,26 @@ def test_opengraph_rejects_empty_source_kind_before_query():
                     "properties": {},
                 }
             ],
-            "edges": [],
+            "edges": [
+                {
+                    "kind": "Jenkins_AdminTo",
+                    "start": {"value": "jenkins-1"},
+                    "end": {"value": "external-1"},
+                    "properties": {},
+                }
+            ],
         },
     }
 
-    with pytest.raises(ValueError, match="metadata.source_kind"):
-        importer._import_opengraph_files_directly(
-            [("jenkins.json", data, "opengraph")]
-        )
+    stub_labels = importer._import_opengraph_files_directly(
+        [("jenkins.json", data, "opengraph")]
+    )
 
-    assert conn.queries == []
+    assert stub_labels == {"OpenGraph_Stub"}
+    assert any(
+        "MERGE (n:`OpenGraph_Stub` {objectid: $objectid})" in q
+        for q, _ in conn.queries
+    )
 
 
 def test_upload_validates_ad_companion_properties_before_filter():