diff --git a/src/config.py b/src/config.py
index 81cf9f9..8363b87 100644
--- a/src/config.py
+++ b/src/config.py
@@ -31,7 +31,7 @@ class Settings(BaseSettings):
     # Processing
     chunk_size: int = Field(800, env="CHUNK_SIZE")                  # tokens per chunk
     chunk_overlap: int = Field(100, env="CHUNK_OVERLAP")
-    quality_threshold: float = Field(0.5, env="QUALITY_THRESHOLD")  # 0-1, chunks below discarded
+    quality_threshold: float = Field(0.3, env="QUALITY_THRESHOLD")  # 0-1, chunks below discarded
 
     # App
     log_level: str = Field("INFO", env="LOG_LEVEL")
diff --git a/src/processor/processor.py b/src/processor/processor.py
index bc4a526..7cb9125 100644
--- a/src/processor/processor.py
+++ b/src/processor/processor.py
@@ -227,6 +227,8 @@ class ContentProcessor:
         import anthropic
         prompt = (
             f'Rate 0-10 how relevant this text is to the topic "{topic}". '
+            f'Be generous — if the text is tangentially related, score 4+. '
+            f'Only score below 3 if completely unrelated. '
             f'Reply with only a number.\n\nText:\n{chunk[:500]}'
         )
         try:
@@ -251,7 +253,7 @@ class ContentProcessor:
                 normalized = min(1.0, score / 10.0)
                 logger.debug("Claude relevance score", raw=score, normalized=round(normalized, 2))
                 return normalized
-            return 0.6
+            return 0.5
         except Exception as e:
             logger.warning("Claude scoring failed, falling back to Ollama", error=str(e))
             return await self._score_with_ollama(chunk, topic)