diff --git a/bigframes/dtypes.py b/bigframes/dtypes.py
index 262fa9dde7..2c5df89665 100644
--- a/bigframes/dtypes.py
+++ b/bigframes/dtypes.py
@@ -754,7 +754,7 @@ def bf_type_from_type_kind(
 
 def is_dtype(scalar: typing.Any, dtype: Dtype) -> bool:
     """Captures whether a scalar can be losslessly represented by a dtype."""
-    if scalar is None:
+    if pd.isna(scalar):
         return True
     if pd.api.types.is_bool_dtype(dtype):
         return pd.api.types.is_bool(scalar)
diff --git a/tests/system/small/test_series.py b/tests/system/small/test_series.py
index 710e1481be..c391370805 100644
--- a/tests/system/small/test_series.py
+++ b/tests/system/small/test_series.py
@@ -629,6 +629,18 @@ def test_series_replace_list_scalar(scalars_dfs):
     )
 
 
+def test_series_replace_nans_with_pd_na(scalars_dfs):
+    scalars_df, scalars_pandas_df = scalars_dfs
+    col_name = "string_col"
+    bf_result = scalars_df[col_name].replace({pd.NA: "UNKNOWN"}).to_pandas()
+    pd_result = scalars_pandas_df[col_name].replace({pd.NA: "UNKNOWN"})
+
+    pd.testing.assert_series_equal(
+        pd_result,
+        bf_result,
+    )
+
+
 @pytest.mark.parametrize(
     ("replacement_dict",),
     (
diff --git a/third_party/bigframes_vendored/ibis/expr/types/core.py b/third_party/bigframes_vendored/ibis/expr/types/core.py
index 9685e4ddca..5704dc993a 100644
--- a/third_party/bigframes_vendored/ibis/expr/types/core.py
+++ b/third_party/bigframes_vendored/ibis/expr/types/core.py
@@ -19,6 +19,7 @@
 import bigframes_vendored.ibis.expr.operations as ops
 from bigframes_vendored.ibis.expr.types.pretty import to_rich
 from bigframes_vendored.ibis.util import experimental
+import pandas as pd
 from public import public
 from rich.console import Console
 from rich.jupyter import JupyterMixin
@@ -34,7 +35,6 @@
         EdgeAttributeGetter,
         NodeAttributeGetter,
     )
-    import pandas as pd
     import polars as pl
     import pyarrow as pa
     import torch
@@ -744,9 +744,9 @@ def _binop(op_class: type[ops.Binary], left: ir.Value, right: ir.Value) -> ir.Va
 
 def _is_null_literal(value: Any) -> bool:
     """Detect whether `value` will be treated by ibis as a null literal."""
-    if value is None:
-        return True
     if isinstance(value, Expr):
         op = value.op()
         return isinstance(op, ops.Literal) and op.value is None
+    if pd.isna(value):
+        return True
     return False