diff --git a/pandas-stubs/core/frame.pyi b/pandas-stubs/core/frame.pyi
index 5f7354578..89d839f9e 100644
--- a/pandas-stubs/core/frame.pyi
+++ b/pandas-stubs/core/frame.pyi
@@ -2106,3 +2106,10 @@ class DataFrame(NDFrame, OpsMixin):
     ) -> DataFrame: ...
     # Move from generic because Series is Generic and it returns Series[bool] there
     def __invert__(self) -> DataFrame: ...
+    def xs(
+        self,
+        key: Hashable,
+        axis: Axis = ...,
+        level: Level | None = ...,
+        drop_level: _bool = ...,
+    ) -> DataFrame | Series: ...
diff --git a/pandas-stubs/core/generic.pyi b/pandas-stubs/core/generic.pyi
index 0cbeb7f54..984a65384 100644
--- a/pandas-stubs/core/generic.pyi
+++ b/pandas-stubs/core/generic.pyi
@@ -15,10 +15,7 @@ from typing import (
 )
 
 import numpy as np
-from pandas import (
-    DataFrame,
-    Index,
-)
+from pandas import Index
 from pandas.core.base import PandasObject
 import pandas.core.indexing as indexing
 import sqlalchemy.engine
@@ -289,13 +286,6 @@ class NDFrame(PandasObject, indexing.IndexingMixin):
     def take(
         self, indices, axis=..., is_copy: _bool | None = ..., **kwargs
     ) -> NDFrame: ...
-    def xs(
-        self,
-        key: Hashable,
-        axis: AxisIndex = ...,
-        level: Level | None = ...,
-        drop_level: _bool = ...,
-    ) -> DataFrame | Series: ...
     def __delitem__(self, idx: Hashable): ...
     def get(self, key: object, default: Dtype | None = ...) -> Dtype: ...
     def reindex_like(
diff --git a/pandas-stubs/core/series.pyi b/pandas-stubs/core/series.pyi
index a514a7195..d95807817 100644
--- a/pandas-stubs/core/series.pyi
+++ b/pandas-stubs/core/series.pyi
@@ -1950,6 +1950,13 @@ class TimedeltaSeries(Series[Timedelta]):
         numeric_only: _bool = ...,
         **kwargs,
     ) -> Timedelta: ...
+    def xs(
+        self,
+        key: Hashable,
+        axis: AxisIndex = ...,
+        level: Level | None = ...,
+        drop_level: _bool = ...,
+    ) -> Series: ...
 
 class PeriodSeries(Series[Period]):
     # ignore needed because of mypy
diff --git a/tests/test_frame.py b/tests/test_frame.py
index d2f7dd030..c0dc5c7f4 100644
--- a/tests/test_frame.py
+++ b/tests/test_frame.py
@@ -2467,3 +2467,19 @@ def test_astype() -> None:
 
     states = pd.DataFrame({"population": population, "area": area})
     check(assert_type(states.astype(object), pd.DataFrame), pd.DataFrame, object)
+
+
+def test_xs_frame_new() -> None:
+    d = {
+        "num_legs": [4, 4, 2, 2],
+        "num_wings": [0, 0, 2, 2],
+        "class": ["mammal", "mammal", "mammal", "bird"],
+        "animal": ["cat", "dog", "bat", "penguin"],
+        "locomotion": ["walks", "walks", "flies", "walks"],
+    }
+    df = pd.DataFrame(data=d)
+    df = df.set_index(["class", "animal", "locomotion"])
+    s1 = df.xs("mammal", axis=0)
+    s2 = df.xs("num_wings", axis=1)
+    check(assert_type(s1, Union[pd.Series, pd.DataFrame]), pd.DataFrame)
+    check(assert_type(s2, Union[pd.Series, pd.DataFrame]), pd.Series)
diff --git a/tests/test_series.py b/tests/test_series.py
index 373431c7f..172e389d4 100644
--- a/tests/test_series.py
+++ b/tests/test_series.py
@@ -1748,3 +1748,9 @@ def test_updated_astype() -> None:
         pd.Series,
         np.integer,
     )
+
+
+def test_check_xs() -> None:
+    s4 = pd.Series([1, 4])
+    s4.xs(0, axis=0)
+    check(assert_type(s4, pd.Series), pd.Series)