diff --git a/hindsight-api-slim/hindsight_api/engine/retain/orchestrator.py b/hindsight-api-slim/hindsight_api/engine/retain/orchestrator.py
index d42559e2b9..ab6ebe894f 100644
--- a/hindsight-api-slim/hindsight_api/engine/retain/orchestrator.py
+++ b/hindsight-api-slim/hindsight_api/engine/retain/orchestrator.py
@@ -549,7 +549,11 @@ async def _extract_and_embed(
     embeddings = await embedding_processing.generate_embeddings_batch(embeddings_model, augmented_texts)
     log_buffer.append(f"  Generate embeddings: {len(embeddings)} embeddings in {time.time() - step_start:.3f}s")
 
-    processed_facts = [ProcessedFact.from_extracted_fact(ef, emb) for ef, emb in zip(extracted_facts, embeddings)]
+    processed_facts = [
+        pf
+        for ef, emb in zip(extracted_facts, embeddings)
+        if (pf := ProcessedFact.from_extracted_fact(ef, emb)) is not None
+    ]
 
     return extracted_facts, processed_facts, chunks, usage
 
diff --git a/hindsight-api-slim/hindsight_api/engine/retain/types.py b/hindsight-api-slim/hindsight_api/engine/retain/types.py
index 85982de7dd..03ef684c10 100644
--- a/hindsight-api-slim/hindsight_api/engine/retain/types.py
+++ b/hindsight-api-slim/hindsight_api/engine/retain/types.py
@@ -5,11 +5,14 @@
 from content input to fact storage.
 """
 
+import logging
 from dataclasses import dataclass, field
 from datetime import datetime
 from typing import Literal, TypedDict
 from uuid import UUID
 
+logger = logging.getLogger(__name__)
+
 
 class RetainContentDict(TypedDict, total=False):
     """Type definition for content items in retain_batch_async.
@@ -187,10 +190,47 @@ def is_duplicate(self) -> bool:
         """Check if this fact was marked as a duplicate."""
         return self.unit_id is None
 
+    @staticmethod
+    def _is_degenerate_text(text: str) -> bool:
+        """Check if fact text has zero information content.
+
+        Rejects empty strings, whitespace-only, single punctuation marks,
+        and common LLM hallucination patterns that carry no semantic meaning.
+        """
+        stripped = (text or "").strip()
+        if not stripped:
+            return True
+        # Single or repeated punctuation patterns with no semantic content
+        degenerate_patterns = {
+            "...",
+            "…",
+            "-",
+            "--",
+            "---",
+            ".",
+            "..",
+            "•",
+            "·",
+            "*",
+            "**",
+            "***",
+            "_,_",
+            "_, _, _",
+        }
+        if stripped in degenerate_patterns:
+            return True
+        # Strings composed entirely of punctuation and whitespace
+        if all(c in ".,;:!?-–—…\"'`´ \t\n\r" for c in stripped):
+            return True
+        # Very short text (<= 2 chars) that is only punctuation
+        if len(stripped) <= 2 and all(not c.isalnum() for c in stripped):
+            return True
+        return False
+
     @staticmethod
     def from_extracted_fact(
         extracted_fact: "ExtractedFact", embedding: list[float], chunk_id: str | None = None
-    ) -> "ProcessedFact":
+    ) -> "ProcessedFact | None":
         """
         Create ProcessedFact from ExtractedFact.
 
@@ -200,8 +240,17 @@ def from_extracted_fact(
             chunk_id: Optional chunk ID
 
         Returns:
-            ProcessedFact ready for storage
+            ProcessedFact ready for storage, or None if the fact text is degenerate
+            (zero information content — punctuation-only, empty, etc.)
         """
+        fact_text = extracted_fact.fact_text or ""
+        if ProcessedFact._is_degenerate_text(fact_text):
+            logger.warning(
+                f"Rejected degenerate fact text: type={extracted_fact.fact_type}, "
+                f"text={fact_text[:80]!r}, entities={extracted_fact.entities}"
+            )
+            return None
+
         # Use occurred dates only if explicitly provided by LLM
         occurred_start = extracted_fact.occurred_start
         occurred_end = extracted_fact.occurred_end
@@ -211,7 +260,7 @@ def from_extracted_fact(
         entities = [EntityRef(name=name) for name in extracted_fact.entities]
 
         return ProcessedFact(
-            fact_text=extracted_fact.fact_text,
+            fact_text=fact_text,
             fact_type=extracted_fact.fact_type,
             embedding=embedding,
             occurred_start=occurred_start,
diff --git a/hindsight-api-slim/hindsight_api/engine/transfer/importer.py b/hindsight-api-slim/hindsight_api/engine/transfer/importer.py
index 92f69ac1bc..cd00a64e83 100644
--- a/hindsight-api-slim/hindsight_api/engine/transfer/importer.py
+++ b/hindsight-api-slim/hindsight_api/engine/transfer/importer.py
@@ -481,7 +481,11 @@ async def _import_one_document(
     if extracted_facts:
         augmented = embedding_processing.augment_texts_with_dates(extracted_facts, format_date_fn)
         embeddings = await embedding_processing.generate_embeddings_batch(embeddings_model, augmented)
-        processed_facts = [ProcessedFact.from_extracted_fact(ef, emb) for ef, emb in zip(extracted_facts, embeddings)]
+        processed_facts = [
+            pf
+            for ef, emb in zip(extracted_facts, embeddings)
+            if (pf := ProcessedFact.from_extracted_fact(ef, emb)) is not None
+        ]
 
     contents = [RetainContent(content=document.original_text or "")]
     chunk_meta = [