diff --git a/src/pipecat/processors/aggregators/llm_response.py b/src/pipecat/processors/aggregators/llm_response.py
index 017973fda..ae0b23507 100644
--- a/src/pipecat/processors/aggregators/llm_response.py
+++ b/src/pipecat/processors/aggregators/llm_response.py
@@ -293,7 +293,13 @@ class LLMUserContextAggregator(LLMContextResponseAggregator):
             await self.push_aggregation()
 
     async def _handle_transcription(self, frame: TranscriptionFrame):
-        self._aggregation += f" {frame.text}" if self._aggregation else frame.text
+        text = frame.text
+
+        # Make sure we really have some text.
+        if not text.strip():
+            return
+
+        self._aggregation += f" {text}" if self._aggregation else text
         # We just got a final result, so let's reset interim results.
         self._seen_interim_results = False
         # Reset aggregation timer.