diff --git a/dask_expr/_shuffle.py b/dask_expr/_shuffle.py
index b01f7cf3..bc3aee6e 100644
--- a/dask_expr/_shuffle.py
+++ b/dask_expr/_shuffle.py
@@ -731,7 +731,9 @@ def _get_index(idx, obj):
                 if idx.ndim == 1:
                     idx = idx.to_frame()
             elif index_shuffle:
-                idx = obj.index.to_frame()
+                # set default index name, otherwise we will end up with 0
+                name = {"name": "_index"} if idx == ["_index"] else {}
+                idx = obj.index.to_frame(**name)
             else:
                 idx = _select_columns_or_index(obj, idx)
             return idx
diff --git a/dask_expr/tests/test_merge.py b/dask_expr/tests/test_merge.py
index 90385e39..7711028a 100644
--- a/dask_expr/tests/test_merge.py
+++ b/dask_expr/tests/test_merge.py
@@ -2,8 +2,9 @@
 
 import numpy as np
 import pytest
+from dask import delayed
 
-from dask_expr import Merge, from_pandas, merge, repartition
+from dask_expr import Merge, from_delayed, from_pandas, merge, repartition
 from dask_expr._expr import Filter, Projection
 from dask_expr._merge import BroadcastJoin
 from dask_expr._shuffle import Shuffle
@@ -1031,6 +1032,22 @@ def test_merge_after_rename(index):
     assert_eq(result, expected, check_index=False)
 
 
+def test_mismatching_meta():
+    df1 = from_pandas(pd.DataFrame({"value": [1, 2, 3]}))
+
+    def creator(i):
+        return pd.DataFrame(
+            {"value2": [1, 2, 3]}, index=pd.Index([1, 2, 3], name="index")
+        )
+
+    df2 = from_delayed(
+        [delayed(creator)(i) for i in range(3)],
+        meta=pd.DataFrame({"value2": [1, 2, 3]}),
+    )
+    result = df1.join(df2)
+    assert_eq(result, df1.compute().join(df2.compute()), check_index=False)
+
+
 def test_merge_tuple_left_on():
     df = pd.DataFrame(
         {