diff --git a/swh/web/api/tests/views/test_origin.py b/swh/web/api/tests/views/test_origin.py
--- a/swh/web/api/tests/views/test_origin.py
+++ b/swh/web/api/tests/views/test_origin.py
@@ -728,6 +728,41 @@
         assert response == expected
 
 
+def test_api_origin_metadata_search_not_in_idx_storage(api_client, mocker):
+    """Tests the origin search for results present in swh-search but not
+    returned by ``origin_intrinsic_metadata_get`` (which happens when results
+    come from extrinsic metadata).
+    """
+
+    mock_idx_storage = mocker.patch("swh.web.utils.archive.idx_storage")
+    mock_idx_storage.origin_intrinsic_metadata_get.return_value = []
+    mock_idx_storage.origin_intrinsic_metadata_search_fulltext.side_effect = (
+        AssertionError("origin_intrinsic_metadata_search_fulltext was called")
+    )
+
+    mock_config = mocker.patch("swh.web.utils.archive.config")
+    mock_config.get_config.return_value = {
+        "search_config": {"metadata_backend": "swh-search"}
+    }
+
+    url = reverse(
+        "api-1-origin-metadata-search",
+        query_params={"fulltext": ORIGIN_METADATA_VALUE},
+    )
+    rv = check_api_get_responses(api_client, url, status_code=200)
+    rv.data = sorted(rv.data, key=lambda d: d["url"])
+
+    expected_data = sorted(
+        [
+            {"url": origin_url, "metadata": {}}
+            for origin_url in sorted(ORIGIN_MASTER_REVISION.keys())
+        ],
+        key=lambda d: d["url"],
+    )
+
+    assert expected_data == rv.data
+
+
 def test_api_origin_metadata_search_limit(api_client, mocker):
     mock_idx_storage = mocker.patch("swh.web.utils.archive.idx_storage")
     oimsft = mock_idx_storage.origin_intrinsic_metadata_search_fulltext
diff --git a/swh/web/utils/archive.py b/swh/web/utils/archive.py
--- a/swh/web/utils/archive.py
+++ b/swh/web/utils/archive.py
@@ -357,18 +357,35 @@
             metadata_pattern=fulltext,
             limit=limit,
         )
-        matches = idx_storage.origin_intrinsic_metadata_get(
-            [r["url"] for r in page_result.results]
-        )
+        origin_urls = [r["url"] for r in page_result.results]
+        metadata = {
+            r.id: r for r in idx_storage.origin_intrinsic_metadata_get(origin_urls)
+        }
+
+        # Results from swh-search are not guaranteed to be in
+        # idx_storage.origin_intrinsic_metadata (typically when they come from
+        # extrinsic metadata; or when the swh-indexer cache is cleared).
+        # When they are missing, we only return the origin url.
+        matches = [
+            metadata[url].to_dict() if url in metadata else {"id": url}
+            for url in origin_urls
+        ]
     else:
-        matches = idx_storage.origin_intrinsic_metadata_search_fulltext(
-            conjunction=[fulltext], limit=limit
-        )
+        matches = [
+            match.to_dict()
+            for match in idx_storage.origin_intrinsic_metadata_search_fulltext(
+                conjunction=[fulltext], limit=limit
+            )
+        ]
 
-    matches = [match.to_dict() for match in matches]
     origins = storage.origin_get([match["id"] for match in matches])
+
     for origin, match in zip(origins, matches):
         if not origin:
+            # filter out origins not present in the storage, as we do not have any
+            # meaningful content to display for that origin at the moment.
+            # This may occur when the storage database we use is lagging behind
+            # swh-search
             continue
         for field in ("from_directory", "from_revision"):
             # from_directory when using swh.indexer >= 2.0.0, from_revision otherwise