diff --git a/crawl4ai/async_url_seeder.py b/crawl4ai/async_url_seeder.py
index 22fa8f630..16fc89997 100644
--- a/crawl4ai/async_url_seeder.py
+++ b/crawl4ai/async_url_seeder.py
@@ -1694,8 +1694,12 @@ def _is_nonsense_url(self, url: str) -> bool:
         if any(pattern in url_lower for pattern in ['?print=', '&print=', '/print/', '_print.']):
             return True
         
-        # 12. Very short paths (likely homepage redirects or errors)
-        if len(path.strip('/')) < 3 and path not in ['/', '/en', '/de', '/fr', '/es', '/it']:
+        # 12. Very short paths (likely homepage redirects or errors).
+        # Compare the slash-stripped path against the whitelist too, so the
+        # canonical trailing-slash form of a language root (e.g. "/en/") is
+        # kept just like "/en" instead of being dropped as nonsense.
+        stripped_path = path.strip('/')
+        if len(stripped_path) < 3 and stripped_path not in ['', 'en', 'de', 'fr', 'es', 'it']:
             return True
         
         return False
diff --git a/tests/unit/test_nonsense_url_language_roots.py b/tests/unit/test_nonsense_url_language_roots.py
new file mode 100644
index 000000000..e3d9ad740
--- /dev/null
+++ b/tests/unit/test_nonsense_url_language_roots.py
@@ -0,0 +1,53 @@
+"""Regression test for ``AsyncUrlSeeder._is_nonsense_url`` language roots.
+
+The "very short path" filter measured length on the slash-stripped path but
+matched the *un-stripped* path against the language-root whitelist, so the
+canonical trailing-slash form (e.g. ``/en/``) was dropped as nonsense even
+though ``/en`` was kept.
+"""
+import sys
+from types import SimpleNamespace
+
+import pytest
+
+# Avoid the optional rank_bm25 dependency at import time (mirrors the sibling
+# unit test); _is_nonsense_url itself does not use BM25.
+sys.modules.setdefault("rank_bm25", SimpleNamespace(BM25Okapi=object))
+
+from crawl4ai.async_url_seeder import AsyncUrlSeeder
+
+
+@pytest.fixture
+def seeder():
+    # _is_nonsense_url only uses its url argument (no instance state), so it is
+    # safe to bypass __init__.
+    return AsyncUrlSeeder.__new__(AsyncUrlSeeder)
+
+
+@pytest.mark.parametrize(
+    "url",
+    [
+        "https://example.com/en",
+        "https://example.com/en/",   # the bug: canonical trailing-slash form
+        "https://example.com/de/",
+        "https://example.com/fr/",
+        "https://example.com/es/",
+        "https://example.com/it/",
+        "https://example.com/",
+        "https://example.com/about",
+    ],
+)
+def test_language_roots_are_not_filtered(seeder, url):
+    assert seeder._is_nonsense_url(url) is False
+
+
+@pytest.mark.parametrize(
+    "url",
+    [
+        "https://example.com/ab/",   # genuinely short, non-language path
+        "https://example.com/x",
+        "https://example.com/robots.txt",
+    ],
+)
+def test_short_junk_and_utility_still_filtered(seeder, url):
+    assert seeder._is_nonsense_url(url) is True