diff --git a/README.md b/README.md
index 6cca3cd0a9..7c683ab5ea 100644
--- a/README.md
+++ b/README.md
@@ -1477,7 +1477,7 @@ previously generated videos. Gemini Enterprise Agent Platform supports video ext
 from google.genai import types
 
 # Read local video (uses mimetypes.guess_type to infer mime type)
-video = types.Video.from_file("local/path/video.mp4")
+video = types.Video.from_file(location="local/path/video.mp4")
 
 # Create operation
 operation = client.models.generate_videos(
diff --git a/docs/source/index.rst b/docs/source/index.rst
index a285fcf808..274a6e82ad 100644
--- a/docs/source/index.rst
+++ b/docs/source/index.rst
@@ -1468,7 +1468,7 @@ Currently, only Gemini Enterprise Agent Platform supports Video to Video generat
     from google.genai import types
 
     # Read local video (uses mimetypes.guess_type to infer mime type)
-    video = types.Video.from_file("local/path/video.mp4")
+    video = types.Video.from_file(location="local/path/video.mp4")
 
     # Create operation
     operation = client.models.generate_videos(
diff --git a/google/genai/tests/readme/__init__.py b/google/genai/tests/readme/__init__.py
new file mode 100644
index 0000000000..66e21cef88
--- /dev/null
+++ b/google/genai/tests/readme/__init__.py
@@ -0,0 +1,16 @@
+# Copyright 2025 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+
diff --git a/google/genai/tests/readme/test_readme_examples.py b/google/genai/tests/readme/test_readme_examples.py
new file mode 100644
index 0000000000..839d25ddd9
--- /dev/null
+++ b/google/genai/tests/readme/test_readme_examples.py
@@ -0,0 +1,177 @@
+# Copyright 2026 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+"""Offline checks that the README examples still match the public SDK.
+
+Running the README needs credentials and network access, so these tests only
+parse its Python blocks (and the copy in docs/source/index.rst) and check, with
+no API call, that every `types.X(...)` and `types.X.method(...)` call is valid
+for the installed SDK: the name exists, keyword arguments are model fields, and
+method calls bind to the real signature.
+"""
+
+import ast
+import inspect
+import pathlib
+import re
+import textwrap
+from typing import Any, Callable
+import unittest
+
+import pydantic
+
+from ... import types
+
+_REPO_ROOT = pathlib.Path(__file__).resolve().parents[4]
+_MARKDOWN_PYTHON_BLOCK = re.compile(r'```python\n(.*?)```', re.DOTALL)
+# An indented reStructuredText block: blank lines or lines with 4+ spaces.
+_RST_PYTHON_BLOCK = re.compile(
+    r'^\.\. code:: python\n\n((?:(?:[ ]{4}.*)?\n)+)', re.MULTILINE
+)
+
+
+def _markdown_blocks(text: str) -> list[str]:
+  return _MARKDOWN_PYTHON_BLOCK.findall(text)
+
+
+def _rst_blocks(text: str) -> list[str]:
+  return [textwrap.dedent(b) for b in _RST_PYTHON_BLOCK.findall(text)]
+
+
+def _types_path(func: ast.expr) -> tuple[str, ...]:
+  """Returns ('Video', 'from_file') for `types.Video.from_file`, else ()."""
+  names = []
+  while isinstance(func, ast.Attribute):
+    names.append(func.attr)
+    func = func.value
+  if isinstance(func, ast.Name) and func.id == 'types' and names:
+    return tuple(reversed(names))
+  return ()
+
+
+def _check_call(name: str, obj: Any, node: ast.Call) -> list[str]:
+  if inspect.isclass(obj) and issubclass(obj, pydantic.BaseModel):
+    fields = set(obj.model_fields)
+    fields.update(f.alias for f in obj.model_fields.values() if f.alias)
+    return [
+        f'{name}() has no field {kw.arg!r}'
+        for kw in node.keywords
+        if kw.arg not in fields
+    ]
+  if not callable(obj):
+    return []
+  try:
+    signature = inspect.signature(obj)
+  except (TypeError, ValueError):
+    return []
+  try:
+    signature.bind(
+        *[None] * len(node.args), **{kw.arg: None for kw in node.keywords}
+    )
+  except TypeError as e:
+    return [f'{name}(): {e}']
+  return []
+
+
+def _find_problems(source: str) -> tuple[int, list[str]]:
+  """Returns (number of calls checked, problems found) for one code block."""
+  try:
+    tree = ast.parse(source)
+  except SyntaxError:
+    return 0, []  # Fragments such as a bare list of contents cannot be parsed.
+  checked = 0
+  problems = []
+  for node in ast.walk(tree):
+    if not isinstance(node, ast.Call):
+      continue
+    if any(isinstance(a, ast.Starred) for a in node.args) or any(
+        kw.arg is None for kw in node.keywords
+    ):
+      continue
+    path = _types_path(node.func)
+    if not path:
+      continue
+    name = '.'.join(('types',) + path)
+    obj: Any = types
+    for attribute in path:
+      obj = getattr(obj, attribute, None)
+      if obj is None:
+        problems.append(f'{name} does not exist')
+        break
+    else:
+      checked += 1
+      problems.extend(_check_call(name, obj, node))
+  return checked, problems
+
+
+class ReadmeExamplesTest(unittest.TestCase):
+
+  def _assert_examples_match_sdk(
+      self, relative_path: str, extract: Callable[[str], list[str]]
+  ) -> None:
+    path = _REPO_ROOT / relative_path
+    if not path.is_file():
+      self.skipTest(f'{relative_path} is not part of this checkout.')
+    checked = 0
+    problems = []
+    for block in extract(path.read_text(encoding='utf-8')):
+      block_checked, block_problems = _find_problems(block)
+      checked += block_checked
+      problems.extend(block_problems)
+    self.assertGreater(checked, 0, f'No examples found in {relative_path}.')
+    self.assertEqual(problems, [], f'Stale examples in {relative_path}.')
+
+  def test_readme_examples_match_sdk(self):
+    self._assert_examples_match_sdk('README.md', _markdown_blocks)
+
+  def test_docs_index_examples_match_sdk(self):
+    self._assert_examples_match_sdk('docs/source/index.rst', _rst_blocks)
+
+  def test_accepts_valid_calls(self):
+    source = (
+        'types.Video.from_file(location="a.mp4")\n'
+        'types.Part.from_text(text="hi")\n'
+        'types.GenerateContentConfig(temperature=0.1)\n'
+    )
+    self.assertEqual(_find_problems(source), (3, []))
+
+  def test_flags_positional_argument_to_keyword_only_classmethod(self):
+    _, problems = _find_problems('types.Video.from_file("a.mp4")')
+    self.assertEqual(len(problems), 1)
+    self.assertIn('types.Video.from_file', problems[0])
+
+  def test_flags_unknown_model_field(self):
+    _, problems = _find_problems('types.GenerateContentConfig(no_such_field=1)')
+    self.assertEqual(
+        problems, ["types.GenerateContentConfig() has no field 'no_such_field'"]
+    )
+
+  def test_flags_unknown_type(self):
+    _, problems = _find_problems('types.NoSuchType(x=1)')
+    self.assertEqual(problems, ['types.NoSuchType does not exist'])
+
+  def test_skips_fragments_that_do_not_parse(self):
+    self.assertEqual(_find_problems('[\n  types.Part(text="a"),\n'), (0, []))
+
+  def test_extracts_rst_code_blocks(self):
+    text = (
+        'Title\n=====\n\n.. code:: python\n\n    x = types.Part()\n\n'
+        '    y = 1\n\nNot code.\n'
+    )
+    self.assertEqual(_rst_blocks(text), ['x = types.Part()\n\ny = 1\n\n'])
+
+
+if __name__ == '__main__':
+  unittest.main()