diff --git a/tests/test_check.py b/tests/test_check.py
index b31973f1..cceff71e 100644
--- a/tests/test_check.py
+++ b/tests/test_check.py
@@ -107,6 +107,31 @@ def test_check_no_description(monkeypatch, capsys):
     )
 
 
+def test_check_invalid_description(monkeypatch, capsys):
+    package = pretend.stub(metadata_dictionary=lambda: {
+        'description': "Project",
+        'long_description': "My amazing project::",
+        'description_content_type': '"text/rst"',  # close, but not quite
+    })
+
+    monkeypatch.setattr(check, "_find_dists", lambda a: ["dist/dist.tar.gz"])
+    monkeypatch.setattr(
+        check,
+        "PackageFile",
+        pretend.stub(from_filename=lambda *a, **kw: package),
+    )
+
+    output_stream = check.StringIO()
+    check.check("dist/*", output_stream=output_stream)
+    assert output_stream.getvalue() == (
+        'Checking distribution dist/dist.tar.gz: '
+        'warning; `long_description_content_type` invalid.\n'
+        'It must be one of the following types: '
+        '[text/markdown, text/plain, text/x-rst].\n'
+        'Passed\n'
+    )
+
+
 def test_check_failing_distribution(monkeypatch):
     renderer = pretend.stub(
         render=pretend.call_recorder(lambda *a, **kw: None)
diff --git a/twine/commands/check.py b/twine/commands/check.py
index 67840bdd..5794905b 100644
--- a/twine/commands/check.py
+++ b/twine/commands/check.py
@@ -32,6 +32,7 @@
     "text/markdown": None,  # Rendering cannot fail
 }
 
+_supported_readme_types = sorted([key for key in _RENDERERS.keys() if key])
 
 # Regular expression used to capture and reformat doctuils warnings into
 # something that a human can understand. This is loosely borrowed from
@@ -88,6 +89,13 @@ def check(dists, output_stream=sys.stdout):
             )
             description_content_type = 'text/x-rst'
 
+        if description_content_type not in _supported_readme_types:
+            output_stream.write(
+                'warning; `long_description_content_type` invalid.\n'
+                'It must be one of the following types: [{}].\n'
+                .format(", ".join(_supported_readme_types))
+            )
+
         content_type, params = cgi.parse_header(description_content_type)
         renderer = _RENDERERS.get(content_type, _RENDERERS[None])