diff --git a/echo/server/dembrane/api/conversation.py b/echo/server/dembrane/api/conversation.py
index 4342fa87..305804cf 100644
--- a/echo/server/dembrane/api/conversation.py
+++ b/echo/server/dembrane/api/conversation.py
@@ -674,6 +674,12 @@ async def summarize_conversation(
     language = conversation_data["project_id"]["language"]
 
     if transcript_str == "":
+        # Update conversation in DB so has_empty_transcript is True, preventing infinite retries by the catch-up scheduler
+        await run_in_thread_pool(
+            conversation_service.update,
+            conversation_id=conversation_id,
+            has_empty_transcript=True,
+        )
         return {
             "status": "success",
             "message": "Transcript is empty, so no summary was generated",
diff --git a/echo/server/dembrane/conversation_utils.py b/echo/server/dembrane/conversation_utils.py
index fb5efa59..836f95e5 100644
--- a/echo/server/dembrane/conversation_utils.py
+++ b/echo/server/dembrane/conversation_utils.py
@@ -134,7 +134,7 @@ def collect_unsummarized_conversations(limit: int = 50) -> List[str]:
     """
     Collect conversations that are fully transcribed but missing a summary.
     
-    Simple check: is_all_chunks_transcribed = True AND summary = null.
+    Simple check: is_all_chunks_transcribed = True AND summary = null AND has_empty_transcript != True.
     The transcribed flag is the source of truth for "ready for summarization".
     
     Args:
@@ -150,6 +150,7 @@ def collect_unsummarized_conversations(limit: int = 50) -> List[str]:
                 "filter": {
                     "is_all_chunks_transcribed": True,
                     "summary": {"_null": True},
+                    "has_empty_transcript": {"_neq": True},
                     "deleted_at": {"_null": True},
                     "created_at": {
                         "_lte": (get_utc_timestamp() - timedelta(minutes=5)).isoformat()
diff --git a/echo/server/dembrane/service/conversation.py b/echo/server/dembrane/service/conversation.py
index 0d53c4e9..a786eca2 100644
--- a/echo/server/dembrane/service/conversation.py
+++ b/echo/server/dembrane/service/conversation.py
@@ -330,6 +330,7 @@ def update(
         is_finished: Any = _UNSET,
         is_all_chunks_transcribed: Any = _UNSET,
         is_over_cap: Any = _UNSET,
+        has_empty_transcript: Any = _UNSET,
     ) -> dict:
         update_data: dict[str, Any] = {}
         if participant_name is not _UNSET:
@@ -348,6 +349,8 @@ def update(
             update_data["is_all_chunks_transcribed"] = is_all_chunks_transcribed
         if is_over_cap is not _UNSET:
             update_data["is_over_cap"] = is_over_cap
+        if has_empty_transcript is not _UNSET:
+            update_data["has_empty_transcript"] = has_empty_transcript
 
         try:
             with self._client_context() as client: