diff --git a/haystack/components/routers/llm_messages_router.py b/haystack/components/routers/llm_messages_router.py
index a47a7771614..498a12fbadc 100644
--- a/haystack/components/routers/llm_messages_router.py
+++ b/haystack/components/routers/llm_messages_router.py
@@ -16,6 +16,15 @@
 _RESERVED_OUTPUT_NAMES = ("chat_generator_text", "unmatched")
 
 
+def _chat_generator_text(generator_result: dict[str, Any]) -> str:
+    """Return the first reply text from a chat generator, or an empty string if none is available."""
+    replies = generator_result.get("replies") or []
+    if not replies:
+        return ""
+    text = getattr(replies[0], "text", None)
+    return text if isinstance(text, str) else ""
+
+
 @component
 class LLMMessagesRouter:
     """
@@ -158,7 +167,7 @@ def run(self, messages: list[ChatMessage]) -> dict[str, str | list[ChatMessage]]
         with _trace_chat_generator_run(self._chat_generator, {"messages": messages_for_inference}) as span:
             generator_result = self._chat_generator.run(messages=messages_for_inference)
             span.set_content_tag("haystack.component.output", generator_result)
-        chat_generator_text = generator_result["replies"][0].text
+        chat_generator_text = _chat_generator_text(generator_result)
 
         output = {"chat_generator_text": chat_generator_text}
 
@@ -207,7 +216,7 @@ async def run_async(self, messages: list[ChatMessage]) -> dict[str, str | list[C
         with _trace_chat_generator_run(self._chat_generator, {"messages": messages_for_inference}) as span:
             generator_result = await _execute_component_async(self._chat_generator, messages=messages_for_inference)
             span.set_content_tag("haystack.component.output", generator_result)
-        chat_generator_text = generator_result["replies"][0].text
+        chat_generator_text = _chat_generator_text(generator_result)
 
         output = {"chat_generator_text": chat_generator_text}
 
diff --git a/releasenotes/notes/fix-llm-messages-router-empty-reply-a9c6d15e4b802f37.yaml b/releasenotes/notes/fix-llm-messages-router-empty-reply-a9c6d15e4b802f37.yaml
new file mode 100644
index 00000000000..22c7e0eb27d
--- /dev/null
+++ b/releasenotes/notes/fix-llm-messages-router-empty-reply-a9c6d15e4b802f37.yaml
@@ -0,0 +1,5 @@
+---
+fixes:
+  - |
+    ``LLMMessagesRouter`` now routes messages to ``unmatched`` when the chat generator returns no replies
+    or a reply with no text, instead of raising ``IndexError`` or ``TypeError``.
diff --git a/test/components/routers/test_llm_messages_router.py b/test/components/routers/test_llm_messages_router.py
index 4feac3fb625..a477f252682 100644
--- a/test/components/routers/test_llm_messages_router.py
+++ b/test/components/routers/test_llm_messages_router.py
@@ -74,6 +74,30 @@ def test_run_input_errors(self):
         with pytest.raises(ValueError):
             router.run([ChatMessage.from_system("You are a helpful assistant.")])
 
+    def test_run_empty_replies_routes_unmatched(self):
+        chat_generator = Mock(spec=["run"])
+        chat_generator.run.return_value = {"replies": []}
+        router = LLMMessagesRouter(
+            chat_generator=chat_generator, output_names=["safe", "unsafe"], output_patterns=["safe", "unsafe"]
+        )
+        messages = [ChatMessage.from_user("Hello")]
+        result = router.run(messages)
+        assert result["chat_generator_text"] == ""
+        assert result["unmatched"] == messages
+        assert "safe" not in result
+        assert "unsafe" not in result
+
+    def test_run_none_reply_text_routes_unmatched(self):
+        chat_generator = Mock(spec=["run"])
+        chat_generator.run.return_value = {"replies": [ChatMessage.from_assistant(text=None)]}
+        router = LLMMessagesRouter(
+            chat_generator=chat_generator, output_names=["safe", "unsafe"], output_patterns=["safe", "unsafe"]
+        )
+        messages = [ChatMessage.from_user("Hello")]
+        result = router.run(messages)
+        assert result["chat_generator_text"] == ""
+        assert result["unmatched"] == messages
+
     def test_run_no_warm_up_with_unwarmable_chat_generator(self):
         chat_generator = Mock(spec=["run"])
         chat_generator.run.return_value = {"replies": [ChatMessage.from_assistant("safe")]}