diff --git a/backends/xnnpack/partition/config/generic_node_configs.py b/backends/xnnpack/partition/config/generic_node_configs.py
index c7a3be5f65f..8a769a8383e 100644
--- a/backends/xnnpack/partition/config/generic_node_configs.py
+++ b/backends/xnnpack/partition/config/generic_node_configs.py
@@ -168,7 +168,8 @@ class AvgPoolingConfig(GenericNodePartitionerConfig):
 
     def check_constraints(self, node: torch.fx.Node, ep: ExportedProgram) -> bool:
         """
-        XNNPACK does not support ceil_mode = True and count_include_pad = True
+        XNNPACK does not support ceil_mode = True, or count_include_pad = True
+        when the node has non-zero padding.
         Additionally, we only support divisor_override if divisor_override = pooling region
         """
         if not self.check_common_constraints(node, ep):
@@ -184,7 +185,7 @@ def check_constraints(self, node: torch.fx.Node, ep: ExportedProgram) -> bool:
         if len(args) >= 6:
             count_include_pad = cast(bool, args[5])
 
-        kernel_size, _, _, _ = normalize_pool2d_args(node, has_dilation=False)
+        kernel_size, _, padding, _ = normalize_pool2d_args(node, has_dilation=False)
         pooling_region = kernel_size[0] * kernel_size[1]
         divisor_override = pooling_region  # Default divisor is pooling_region
         if len(args) >= 7:
@@ -194,7 +195,7 @@ def check_constraints(self, node: torch.fx.Node, ep: ExportedProgram) -> bool:
             why(node, reason="ceil mode is not supported")
             return False
 
-        if count_include_pad:
+        if count_include_pad and any(p != 0 for p in padding):
             why(
                 node,
                 reason="zero-padding in the averaging calculation is not supported",
diff --git a/backends/xnnpack/test/ops/test_avgpool2d.py b/backends/xnnpack/test/ops/test_avgpool2d.py
index 23ca5f20e5a..909569daf8e 100644
--- a/backends/xnnpack/test/ops/test_avgpool2d.py
+++ b/backends/xnnpack/test/ops/test_avgpool2d.py
@@ -67,7 +67,8 @@ def test_fp32_avgpool2d_ceil_mode_unsupported(self):
 
     def test_fp32_avgpool2d_count_include_pad_unsupported(self):
         """
-        The XNNPACK backend does not support count_include_pad=True.
+        The XNNPACK backend does not support count_include_pad=True with
+        non-zero padding.
         """
         inputs = (torch.randn(1, 1, 10, 10),)
         (
@@ -78,6 +79,32 @@ def test_fp32_avgpool2d_count_include_pad_unsupported(self):
             .check_not(["torch.ops.higher_order.executorch_call_delegate"])
         )
 
+    class AvgPool2dAllDefaults(torch.nn.Module):
+        def __init__(self):
+            super().__init__()
+            self.avgPool = torch.nn.AvgPool2d(2)
+
+        def forward(self, x):
+            return self.avgPool(x)
+
+    def test_fp32_avgpool2d_default_args(self):
+        """
+        count_include_pad defaults to True, but with the default zero padding
+        there are no padded elements to count, so the node must still delegate.
+        """
+        inputs = (torch.randn(1, 1, 10, 10),)
+        (
+            Tester(self.AvgPool2dAllDefaults(), inputs)
+            .export()
+            .check_count({"torch.ops.aten.avg_pool2d.default": 1})
+            .to_edge_transform_and_lower()
+            .check_count({"torch.ops.higher_order.executorch_call_delegate": 1})
+            .check_not(["executorch_exir_dialects_edge__ops_aten_avg_pool2d_default"])
+            .to_executorch()
+            .serialize()
+            .run_method_and_compare_outputs()
+        )
+
     class AvgPool2dSingleElementKernel(torch.nn.Module):
         def __init__(self, divisor_override=None):
             super().__init__()