diff --git a/xarray/core/concat.py b/xarray/core/concat.py
index b23073e7e70..f4af256430c 100644
--- a/xarray/core/concat.py
+++ b/xarray/core/concat.py
@@ -532,7 +532,7 @@ def get_indexes(name):
             elif name == dim:
                 var = ds._variables[name]
                 if not var.dims:
-                    yield PandasIndex([var.values.item()], dim)
+                    yield PandasIndex([var.values.item()], dim, coord_dtype=var.dtype)
 
     # stack up each variable and/or index to fill-out the dataset (in order)
     # n.b. this loop preserves variable order, needed for groupby.
diff --git a/xarray/tests/test_concat.py b/xarray/tests/test_concat.py
index f31b6972187..c0b98e28b12 100644
--- a/xarray/tests/test_concat.py
+++ b/xarray/tests/test_concat.py
@@ -456,6 +456,17 @@ def test_concat_promote_shape(self) -> None:
         )
         assert_identical(actual, expected)
 
+        # regression GH6416 (coord dtype)
+        time_data1 = np.array(["2022-01-01", "2022-02-01"], dtype="datetime64[ns]")
+        time_data2 = np.array("2022-03-01", dtype="datetime64[ns]")
+        time_expected = np.array(
+            ["2022-01-01", "2022-02-01", "2022-03-01"], dtype="datetime64[ns]"
+        )
+        objs = [Dataset({}, {"time": time_data1}), Dataset({}, {"time": time_data2})]
+        actual = concat(objs, "time")
+        expected = Dataset({}, {"time": time_expected})
+        assert_identical(actual, expected)
+
     def test_concat_do_not_promote(self) -> None:
         # GH438
         objs = [