diff --git a/src/crawlee/_utils/requests.py b/src/crawlee/_utils/requests.py
index fa31d4621d..3fe2ee7fd9 100644
--- a/src/crawlee/_utils/requests.py
+++ b/src/crawlee/_utils/requests.py
@@ -17,10 +17,11 @@ def normalize_url(url: str, *, keep_url_fragment: bool = False) -> str:
     """Normalize a URL.
 
     This function cleans and standardizes a URL by removing leading and trailing whitespaces,
-    converting the scheme and netloc to lower case, stripping unwanted tracking parameters
+    converting the scheme and host to lower case, stripping unwanted tracking parameters
     (specifically those beginning with 'utm_'), sorting the remaining query parameters alphabetically,
-    and optionally retaining the URL fragment. The goal is to ensure that URLs that are functionally
-    identical but differ in trivial ways (such as parameter order or casing) are treated as the same.
+    and optionally retaining the URL fragment. Per RFC 3986 only the scheme and host are
+    case-insensitive: the path, query, and fragment keep their original casing, so case-distinct
+    pages are not silently deduplicated.
 
     Args:
         url: The URL to be normalized.
@@ -29,7 +30,8 @@ def normalize_url(url: str, *, keep_url_fragment: bool = False) -> str:
     Returns:
         A string containing the normalized URL.
     """
-    # Parse the URL
+    # Parse the URL. yarl already lowercases the scheme and host per RFC 3986 while
+    # preserving the casing of the path, query, and fragment.
     parsed_url = URL(url.strip())
 
     # Remove any 'utm_' parameters
@@ -44,7 +46,7 @@ def normalize_url(url: str, *, keep_url_fragment: bool = False) -> str:
         yarl_new_url.path.removesuffix('/'), keep_query=True, keep_fragment=keep_url_fragment
     )
 
-    return str(yarl_new_url).lower()
+    return str(yarl_new_url)
 
 
 def compute_unique_key(
diff --git a/tests/unit/_utils/test_requests.py b/tests/unit/_utils/test_requests.py
index 8198909592..033b0bab4f 100644
--- a/tests/unit/_utils/test_requests.py
+++ b/tests/unit/_utils/test_requests.py
@@ -15,22 +15,37 @@
             'http://example.com/?another_key=another_value&key=value',
             False,
         ),
-        ('HTTPS://EXAMPLE.COM/?KEY=VALUE', 'https://example.com/?key=value', False),
+        ('HTTPS://EXAMPLE.COM/?KEY=VALUE', 'https://example.com/?KEY=VALUE', False),
         ('', '', False),
         ('http://example.com/#fragment', 'http://example.com/#fragment', True),
         ('http://example.com/#fragment', 'http://example.com', False),
         ('  https://example.com/  ', 'https://example.com', False),
         ('http://example.com/?b=2&a=1', 'http://example.com/?a=1&b=2', False),
+        # RFC 3986: only the scheme and host are case-insensitive. Path, query,
+        # and fragment casing is significant and must survive normalization, so
+        # case-distinct pages are not silently deduplicated (#2008).
+        ('https://example.com/Product/ABC', 'https://example.com/Product/ABC', False),
+        ('https://example.com/Product/ABC/', 'https://example.com/Product/ABC', False),
+        ('https://example.com/?token=SeCrEt', 'https://example.com/?token=SeCrEt', False),
+        (
+            'https://example.com/Path?Token=SeCrEt#Frag',
+            'https://example.com/Path?Token=SeCrEt#Frag',
+            True,
+        ),
     ],
     ids=[
         'remove_utm_params',
         'retain_sort_non_utm_params',
-        'convert_scheme_netloc_to_lowercase',
+        'lowercase_scheme_and_host_only',
         'handle_empty_url',
         'retain_fragment',
         'remove_fragment',
         'trim_whitespace',
         'sort_query_params',
+        'preserve_path_case',
+        'preserve_path_case_with_trailing_slash',
+        'preserve_query_value_case',
+        'preserve_path_query_fragment_case',
     ],
 )
 def test_normalize_url(url: str, expected_output: str, *, keep_url_fragment: bool) -> None:
@@ -38,6 +53,16 @@ def test_normalize_url(url: str, expected_output: str, *, keep_url_fragment: boo
     assert output == expected_output
 
 
+def test_normalize_url_does_not_collapse_case_distinct_paths() -> None:
+    # Two URLs differing only in path casing must normalize to different URLs,
+    # so compute_unique_key does not silently deduplicate case-distinct pages.
+    upper = normalize_url('https://example.com/Product/ABC')
+    lower = normalize_url('https://example.com/product/abc')
+    assert upper != lower
+    assert upper == 'https://example.com/Product/ABC'
+    assert lower == 'https://example.com/product/abc'
+
+
 def test_compute_unique_key_basic() -> None:
     url = 'https://crawlee.dev'
     uk_get = compute_unique_key(url, method='GET')