diff --git a/litellm/integrations/otel/model/payloads.py b/litellm/integrations/otel/model/payloads.py
index 484f4a4c294..c23b3291365 100644
--- a/litellm/integrations/otel/model/payloads.py
+++ b/litellm/integrations/otel/model/payloads.py
@@ -720,6 +720,7 @@ class _ToolCall(TypedDict):
 class _AssistantMessage(TypedDict):
     role: ReadOnly[str]
     content: ReadOnly[str | None]
+    refusal: ReadOnly[str | None]
     tool_calls: ReadOnly[tuple[_ToolCall, ...] | None]
 
 
@@ -735,13 +736,7 @@ def _responses_choices(response: Mapping[str, object]) -> tuple[_Choice, ...]:
     """A Responses API ``output`` folded into one chat-shaped assistant choice."""
     items: Final = _dicts(response.get("output"))
     messages: Final = tuple(item for item in items if item.get("type") == "message")
-    content: Final = "".join(
-        text
-        for item in messages
-        for part in _dicts(item.get("content"))
-        if part.get("type") == "output_text"
-        if (text := as_str(part.get("text"))) is not None
-    )
+    parts: Final = tuple(part for item in messages for part in _dicts(item.get("content")))
     tool_calls: Final = tuple(
         _responses_tool_call(item) for item in items if item.get("type") in _RESPONSES_TOOL_CALL_TYPES
     )
@@ -749,13 +744,21 @@ def _responses_choices(response: Mapping[str, object]) -> tuple[_Choice, ...]:
         return ()
     message: Final[_AssistantMessage] = {
         "role": next((role for item in messages if (role := as_str(item.get("role")))), "assistant"),
-        "content": content if messages else None,
+        "content": _responses_parts_text(parts, "output_text", "text"),
+        "refusal": _responses_parts_text(parts, "refusal", "refusal"),
         "tool_calls": tool_calls or None,
     }
     choice: Final[_Choice] = {"message": message, "finish_reason": _responses_finish_reason(response, bool(tool_calls))}
     return (choice,)
 
 
+def _responses_parts_text(parts: tuple[Mapping[str, object], ...], part_type: str, field: str) -> str | None:
+    texts: Final = tuple(
+        text for part in parts if part.get("type") == part_type if (text := as_str(part.get(field))) is not None
+    )
+    return "".join(texts) if texts else None
+
+
 def _responses_tool_call(item: Mapping[str, object]) -> _ToolCall:
     custom: Final = item.get("type") == "custom_tool_call"
     function: Final[_ToolFunction] = {
diff --git a/tests/test_litellm/integrations/otel/test_otel_v2_sources_of_truth.py b/tests/test_litellm/integrations/otel/test_otel_v2_sources_of_truth.py
index 972c91670f8..17de3cf1e8a 100644
--- a/tests/test_litellm/integrations/otel/test_otel_v2_sources_of_truth.py
+++ b/tests/test_litellm/integrations/otel/test_otel_v2_sources_of_truth.py
@@ -761,7 +761,7 @@ def test_responses_output_text_becomes_one_assistant_choice_with_stop():
 
     assert json.loads(json.dumps(data.choices_out)) == [
         {
-            "message": {"role": "assistant", "content": "pong", "tool_calls": None},
+            "message": {"role": "assistant", "content": "pong", "refusal": None, "tool_calls": None},
             "finish_reason": "stop",
         }
     ]
@@ -835,6 +835,23 @@ def test_responses_content_only_reads_output_text_parts():
     data = LLMCallSpanData.from_standard_logging_payload(_responses_payload([item]), capture_content=True)
 
     assert data.choices_out[0]["message"]["content"] == "ok"
+    assert data.choices_out[0]["message"]["refusal"] == "no"
+
+
+def test_responses_refusal_only_output_keeps_the_refusal_text():
+    item = {
+        "type": "message",
+        "role": "assistant",
+        "content": [{"type": "refusal", "refusal": "I can't "}, {"type": "refusal", "refusal": "help with that."}],
+    }
+    data = LLMCallSpanData.from_standard_logging_payload(_responses_payload([item]), capture_content=True)
+
+    assert json.loads(json.dumps(data.choices_out)) == [
+        {
+            "message": {"role": "assistant", "content": None, "refusal": "I can't help with that.", "tool_calls": None},
+            "finish_reason": "stop",
+        }
+    ]
 
 
 def test_responses_output_without_messages_or_tool_calls_stays_empty():
diff --git a/tests/test_litellm/integrations/otel/test_otel_v2_vendor_mappers.py b/tests/test_litellm/integrations/otel/test_otel_v2_vendor_mappers.py
index 5b4d1e7a802..4e375de0494 100644
--- a/tests/test_litellm/integrations/otel/test_otel_v2_vendor_mappers.py
+++ b/tests/test_litellm/integrations/otel/test_otel_v2_vendor_mappers.py
@@ -218,6 +218,7 @@ def test_langfuse_mapper_renders_a_responses_api_call_from_the_standard_logging_
         {
             "role": "assistant",
             "content": "Checking.",
+            "refusal": None,
             "tool_calls": [
                 {"id": "call_1", "type": "function", "function": {"name": "get_weather", "arguments": '{"city": "sf"}'}}
             ],