diff --git a/bigframes/session/__init__.py b/bigframes/session/__init__.py
index 2bc612bdbe..40831292de 100644
--- a/bigframes/session/__init__.py
+++ b/bigframes/session/__init__.py
@@ -916,6 +916,12 @@ def read_pandas(self, pandas_dataframe: pandas.DataFrame) -> dataframe.DataFrame
     def _read_pandas(
         self, pandas_dataframe: pandas.DataFrame, api_name: str
     ) -> dataframe.DataFrame:
+        if isinstance(pandas_dataframe, dataframe.DataFrame):
+            raise ValueError(
+                "read_pandas() expects a pandas.DataFrame, but got a "
+                "bigframes.pandas.DataFrame."
+            )
+
         if (
             pandas_dataframe.size < MAX_INLINE_DF_SIZE
             # TODO(swast): Workaround data types limitation in inline data.
diff --git a/tests/unit/session/test_io_pandas.py b/tests/unit/session/test_io_pandas.py
index 959cccd80e..81d02466ef 100644
--- a/tests/unit/session/test_io_pandas.py
+++ b/tests/unit/session/test_io_pandas.py
@@ -13,7 +13,9 @@
 # limitations under the License.
 
 import datetime
+import re
 from typing import Dict, Union
+import unittest.mock as mock
 
 import geopandas  # type: ignore
 import numpy
@@ -24,8 +26,11 @@
 import pytest
 
 import bigframes.features
+import bigframes.pandas
 import bigframes.session._io.pandas
 
+from .. import resources
+
 _LIST_OF_SCALARS = [
     [1, 2, 3],
     [],
@@ -475,3 +480,13 @@ def test_arrow_to_pandas_wrong_size_dtypes(
 ):
     with pytest.raises(ValueError, match=f"Number of types {len(dtypes)}"):
         bigframes.session._io.pandas.arrow_to_pandas(arrow_table, dtypes)
+
+
+def test_read_pandas_with_bigframes_dataframe():
+    session = resources.create_bigquery_session()
+    df = mock.create_autospec(bigframes.pandas.DataFrame, instance=True)
+
+    with pytest.raises(
+        ValueError, match=re.escape("read_pandas() expects a pandas.DataFrame")
+    ):
+        session.read_pandas(df)