diff --git a/seaborn/axisgrid.py b/seaborn/axisgrid.py
index 17d333bc89..6c7fe4de77 100644
--- a/seaborn/axisgrid.py
+++ b/seaborn/axisgrid.py
@@ -1248,12 +1248,43 @@ def __init__(
         data = handle_data_source(data)
 
         # Sort out the variables that define the grid
-        numeric_cols = self._find_numeric_cols(data)
-        if hue in numeric_cols:
-            numeric_cols.remove(hue)
-        if vars is not None:
+
+        # `_find_numeric_cols` will crash if there are duplicated columns in the data,
+        # even if these columns are not in the `vars` variable.
+        # If `vars` is provided, we don't need `_find_numeric_cols`.
+        # If `vars` is not provided, we need `_find_numeric_cols`,
+        # but it crashes with ambigious ValueError: The truth value of a DataFrame ...
+        # My fix is to skip `_find_numeric_cols` when `vars` is provided.
+        # And raise early error if data_to_plot.columns is duplicated.
+        # [I suppose duplicated columns are not expected in PairGrid]
+
+        if vars is not None:    # user provide vars
             x_vars = list(vars)
             y_vars = list(vars)
+            if len(set(vars)) < len(x_vars):
+                # Does not crash, only causes unexpected figures.
+                # Do not take efforts to specify duplicants.
+                warnings.warn(f"Duplicated items in vars: {x_vars}")
+                condensed_vars = list(set(x_vars))
+            else:
+                condensed_vars = x_vars
+                # Use condensed_vars to avoid duplicated items in vars
+                # causing duplicates in data.loc[:, vars].columns.
+            selected_columns = data.loc[:, condensed_vars].columns
+            if not selected_columns.is_unique:
+                # Crash if duplicated columns are selected in vars.
+                # Specify duplicants since we raise an Error.
+                dupe_cols = selected_columns[selected_columns.duplicated()]
+                raise ValueError(
+                    f"Columns: {dupe_cols} are duplicated.")
+        else:
+            if not data.columns.is_unique:
+                dupe_cols = data.columns[data.columns.duplicated()]
+                raise ValueError(
+                    f"Columns: {dupe_cols} are duplicated.")
+            numeric_cols = self._find_numeric_cols(data)
+            if hue in numeric_cols:
+                numeric_cols.remove(hue)
         if x_vars is None:
             x_vars = numeric_cols
         if y_vars is None:
diff --git a/tests/test_axisgrid.py b/tests/test_axisgrid.py
index 6470edfa4f..d53321afd2 100644
--- a/tests/test_axisgrid.py
+++ b/tests/test_axisgrid.py
@@ -785,6 +785,32 @@ def test_remove_hue_from_default(self):
         assert hue in g.x_vars
         assert hue in g.y_vars
 
+    def test_duplicates_in_df_columns_without_vars(self):
+        # should fail with clear msg
+        df_with_dupe = self.df.loc[:, ["x", "y", "z"]].copy()
+        df_with_dupe.columns = ["x", "y", "y"]
+        with pytest.raises(ValueError, match=r"Columns: .* are duplicated\."):
+            ag.PairGrid(df_with_dupe)
+
+    def test_duplicates_in_df_columns_with_related_vars(self):
+        # should fail with clear msg
+        df_with_dupe = self.df.loc[:, ["x", "y", "z"]].copy()
+        df_with_dupe.columns = ["x", "y", "y"]
+        with pytest.raises(ValueError, match=r"Columns: .* are duplicated\."):
+            ag.PairGrid(df_with_dupe, vars=['x', 'y'])
+
+    def test_duplicated_vars(self):
+        # should only warn
+        with pytest.warns(UserWarning, match=r"Duplicated items in vars: .*"):
+            ag.PairGrid(self.df, vars=['x', 'y', 'y'])
+
+    def test_duplicates_in_df_columns_with_not_related_vars(self):
+        # should pass
+        df_with_dupe = pd.concat(
+            [self.df["x"], self.df["y"], self.df["z"], self.df["x"]], axis=1)
+        df_with_dupe.columns = ["x", "y", "z", "z"]
+        ag.PairGrid(df_with_dupe, vars=['x', 'y'])
+
     @pytest.mark.parametrize(
         "x_vars, y_vars",
         [