diff --git a/src/google/adk/telemetry.py b/src/google/adk/telemetry.py
index a09c2f55b..10ac58399 100644
--- a/src/google/adk/telemetry.py
+++ b/src/google/adk/telemetry.py
@@ -202,7 +202,7 @@ def trace_call_llm(
     )
     span.set_attribute(
         'gen_ai.usage.output_tokens',
-        llm_response.usage_metadata.total_token_count,
+        llm_response.usage_metadata.candidates_token_count,
     )
 
 
diff --git a/tests/unittests/test_telemetry.py b/tests/unittests/test_telemetry.py
index cf115d5f0..8a3964b21 100644
--- a/tests/unittests/test_telemetry.py
+++ b/tests/unittests/test_telemetry.py
@@ -155,7 +155,9 @@ async def test_trace_call_llm_usage_metadata(monkeypatch, mock_span_fixture):
   llm_response = LlmResponse(
       turn_complete=True,
       usage_metadata=types.GenerateContentResponseUsageMetadata(
-          total_token_count=100, prompt_token_count=50
+          total_token_count=100,
+          prompt_token_count=50,
+          candidates_token_count=50,
       ),
   )
   trace_call_llm(invocation_context, 'test_event_id', llm_request, llm_response)
@@ -163,7 +165,7 @@ async def test_trace_call_llm_usage_metadata(monkeypatch, mock_span_fixture):
   expected_calls = [
       mock.call('gen_ai.system', 'gcp.vertex.agent'),
       mock.call('gen_ai.usage.input_tokens', 50),
-      mock.call('gen_ai.usage.output_tokens', 100),
+      mock.call('gen_ai.usage.output_tokens', 50),
   ]
   assert mock_span_fixture.set_attribute.call_count == 9
   mock_span_fixture.set_attribute.assert_has_calls(