diff --git a/tests/test_url.py b/tests/test_url.py
index 784afd2..92a5bc3 100644
--- a/tests/test_url.py
+++ b/tests/test_url.py
@@ -285,3 +285,20 @@ def test_embed() -> None:
 def test_pchar() -> None:
     url = URL("s3://mybucket") / "some_folder/123_2017-10-30T18:43:11.csv.gz"
     assert str(url) == "s3://mybucket/some_folder/123_2017-10-30T18:43:11.csv.gz"
+
+
+def test_percent_encoding_spaces() -> None:
+    """Test that %20 encoded spaces don't get double-encoded."""
+    # Reported bug: URL with %20 in middle of path segment gets double-encoded to %2520
+    url = URL("https://somepath.com/test/Test%20path/my%20test%20file.txt")
+    assert str(url) == "https://somepath.com/test/Test%20path/my%20test%20file.txt"
+
+    # Test various positions of %20
+    assert str(URL("https://somepath.com/Test%20path")) == "https://somepath.com/Test%20path"
+    assert str(URL("https://somepath.com/%20leading")) == "https://somepath.com/%20leading"
+    assert str(URL("https://somepath.com/trailing%20")) == "https://somepath.com/trailing%20"
+    assert str(URL("https://somepath.com/multiple%20spaces%20here")) == "https://somepath.com/multiple%20spaces%20here"
+
+    # Test that actual spaces get encoded properly
+    url_with_spaces = URL("https://somepath.com/test") / "Test path" / "my test file.txt"
+    assert str(url_with_spaces) == "https://somepath.com/test/Test%20path/my%20test%20file.txt"
diff --git a/urlpath/__init__.py b/urlpath/__init__.py
index e5de855..e16365a 100644
--- a/urlpath/__init__.py
+++ b/urlpath/__init__.py
@@ -470,11 +470,12 @@ def path(self) -> str:
 
         begin = 1 if self._drv or self._root else 0
 
+        # Decode parts before encoding to avoid double-encoding
+        parts = [urllib.parse.unquote(i) for i in self._parts[begin:-1]] + [self.name]
+
         return (
             self._root
-            + self._flavour.sep.join(
-                urllib.parse.quote(i, safe=safe_pchars) for i in self._parts[begin:-1] + [self.name]
-            )
+            + self._flavour.sep.join(urllib.parse.quote(i, safe=safe_pchars) for i in parts)
             + self.trailing_sep
         )