diff --git a/narwhals/functions.py b/narwhals/functions.py
index 95c26a0c13..cbeecc0152 100644
--- a/narwhals/functions.py
+++ b/narwhals/functions.py
@@ -591,20 +591,33 @@ def read_csv(
         |     1  2   5     |
         └──────────────────┘
     """
-    eager_backend = Implementation.from_backend(backend)
-    native_namespace = eager_backend.to_native_namespace()
+    impl = Implementation.from_backend(backend)
+    native_namespace = impl.to_native_namespace()
     native_frame: NativeFrame
-    if eager_backend in {
+    if impl in {
         Implementation.POLARS,
         Implementation.PANDAS,
         Implementation.MODIN,
         Implementation.CUDF,
     }:
         native_frame = native_namespace.read_csv(source, **kwargs)
-    elif eager_backend is Implementation.PYARROW:
+    elif impl is Implementation.PYARROW:
         from pyarrow import csv  # ignore-banned-import
 
         native_frame = csv.read_csv(source, **kwargs)
+    elif impl in {
+        Implementation.PYSPARK,
+        Implementation.DASK,
+        Implementation.DUCKDB,
+        Implementation.IBIS,
+        Implementation.SQLFRAME,
+        Implementation.PYSPARK_CONNECT,
+    }:
+        msg = (
+            f"Expected eager backend, found {impl}.\n\n"
+            f"Hint: use nw.scan_csv(source={source}, backend={backend})"
+        )
+        raise ValueError(msg)
     else:  # pragma: no cover
         try:
             # implementation is UNKNOWN, Narwhals extension using this feature should
@@ -734,22 +747,33 @@ def read_parquet(
         |c: [[0.2,0.1]]    |
         └──────────────────┘
     """
-    implementation = Implementation.from_backend(backend)
-    native_namespace = implementation.to_native_namespace()
+    impl = Implementation.from_backend(backend)
+    native_namespace = impl.to_native_namespace()
     native_frame: NativeFrame
-    if implementation in {
+    if impl in {
         Implementation.POLARS,
         Implementation.PANDAS,
         Implementation.MODIN,
         Implementation.CUDF,
-        Implementation.DUCKDB,
-        Implementation.IBIS,
     }:
         native_frame = native_namespace.read_parquet(source, **kwargs)
-    elif implementation is Implementation.PYARROW:
+    elif impl is Implementation.PYARROW:
         import pyarrow.parquet as pq  # ignore-banned-import
 
         native_frame = pq.read_table(source, **kwargs)
+    elif impl in {
+        Implementation.PYSPARK,
+        Implementation.DASK,
+        Implementation.DUCKDB,
+        Implementation.IBIS,
+        Implementation.SQLFRAME,
+        Implementation.PYSPARK_CONNECT,
+    }:
+        msg = (
+            f"Expected eager backend, found {impl}.\n\n"
+            f"Hint: use nw.scan_parquet(source={source}, backend={backend})"
+        )
+        raise ValueError(msg)
     else:  # pragma: no cover
         try:
             # implementation is UNKNOWN, Narwhals extension using this feature should
diff --git a/tests/read_scan_test.py b/tests/read_scan_test.py
index 737f83953e..349f71cbb8 100644
--- a/tests/read_scan_test.py
+++ b/tests/read_scan_test.py
@@ -46,6 +46,17 @@ def test_read_csv_kwargs(tmpdir: pytest.TempdirFactory) -> None:
     assert_equal_data(result, data)
 
 
+@pytest.mark.parametrize("backend", ["duckdb", "ibis", "sqlframe"])
+def test_read_csv_raise_with_lazy(tmpdir: pytest.TempdirFactory, backend: str) -> None:
+    pytest.importorskip(backend)
+    df_pl = pl.DataFrame(data)
+    filepath = str(tmpdir / "file.csv")  # type: ignore[operator]
+    df_pl.write_csv(filepath)
+
+    with pytest.raises(ValueError, match="Expected eager backend, found"):
+        nw.read_csv(filepath, backend=backend)
+
+
 def test_scan_csv(tmpdir: pytest.TempdirFactory, constructor: Constructor) -> None:
     kwargs: dict[str, Any]
     if "sqlframe" in str(constructor):
@@ -119,6 +130,19 @@ def test_read_parquet_kwargs(tmpdir: pytest.TempdirFactory) -> None:
     assert_equal_data(result, data)
 
 
+@pytest.mark.parametrize("backend", ["duckdb", "ibis", "sqlframe"])
+def test_read_parquet_raise_with_lazy(
+    tmpdir: pytest.TempdirFactory, backend: str
+) -> None:
+    pytest.importorskip(backend)
+    df_pl = pl.DataFrame(data)
+    filepath = str(tmpdir / "file.parquet")  # type: ignore[operator]
+    df_pl.write_parquet(filepath)
+
+    with pytest.raises(ValueError, match="Expected eager backend, found"):
+        nw.read_parquet(filepath, backend=backend)
+
+
 @pytest.mark.skipif(PANDAS_VERSION < (1, 5), reason="too old for pyarrow")
 def test_scan_parquet(tmpdir: pytest.TempdirFactory, constructor: Constructor) -> None:
     kwargs: dict[str, Any]