diff --git a/changelog/3888.fixed.md b/changelog/3888.fixed.md
new file mode 100644
index 000000000..99e9ad0e0
--- /dev/null
+++ b/changelog/3888.fixed.md
@@ -0,0 +1 @@
+- Fixed turn completion instructions being lost when `LLMMessagesUpdateFrame` replaces the LLM context. When `filter_incomplete_user_turns` is enabled, the turn completion system message is now re-injected after context replacement.
diff --git a/src/pipecat/processors/aggregators/llm_response_universal.py b/src/pipecat/processors/aggregators/llm_response_universal.py
index c43cc279d..96f3702be 100644
--- a/src/pipecat/processors/aggregators/llm_response_universal.py
+++ b/src/pipecat/processors/aggregators/llm_response_universal.py
@@ -642,6 +642,9 @@ class LLMUserAggregator(LLMContextAggregator):
 
     async def _handle_llm_messages_update(self, frame: LLMMessagesUpdateFrame):
         self.set_messages(frame.messages)
+        if self._params.filter_incomplete_user_turns:
+            config = self._params.user_turn_completion_config or UserTurnCompletionConfig()
+            self._context.add_message({"role": "system", "content": config.completion_instructions})
         if frame.run_llm:
             await self.push_context_frame()
 
diff --git a/tests/test_context_aggregators_universal.py b/tests/test_context_aggregators_universal.py
index e86905e1c..b22abf6c6 100644
--- a/tests/test_context_aggregators_universal.py
+++ b/tests/test_context_aggregators_universal.py
@@ -50,6 +50,7 @@ from pipecat.turns.user_mute import (
     MuteUntilFirstBotCompleteUserMuteStrategy,
 )
 from pipecat.turns.user_stop import SpeechTimeoutUserTurnStopStrategy
+from pipecat.turns.user_turn_completion_mixin import UserTurnCompletionConfig
 from pipecat.turns.user_turn_strategies import UserTurnStrategies
 
 USER_TURN_STOP_TIMEOUT = 0.2
@@ -155,6 +156,28 @@ class TestLLMUserAggregator(unittest.IsolatedAsyncioTestCase):
         )
         assert context.messages[0]["content"] == "Hi there!"
 
+    async def test_llm_messages_update_reinjects_turn_completion_instructions(self):
+        context = LLMContext()
+        params = LLMUserAggregatorParams(filter_incomplete_user_turns=True)
+        pipeline = Pipeline([LLMUserAggregator(context, params=params)])
+
+        new_messages = [
+            {"role": "system", "content": "You are a helpful assistant."},
+            {"role": "user", "content": "Hello!"},
+        ]
+        frames_to_send = [LLMMessagesUpdateFrame(messages=new_messages)]
+        await run_test(
+            pipeline,
+            frames_to_send=frames_to_send,
+        )
+        config = UserTurnCompletionConfig()
+        # The context should contain the new messages plus the re-injected instructions
+        assert len(context.messages) == 3
+        assert context.messages[0]["content"] == "You are a helpful assistant."
+        assert context.messages[1]["content"] == "Hello!"
+        assert context.messages[2]["role"] == "system"
+        assert context.messages[2]["content"] == config.completion_instructions
+
     async def test_default_user_turn_strategies(self):
         context = LLMContext()
         user_aggregator = LLMUserAggregator(