diff --git a/pyproject.toml b/pyproject.toml
index e3418ab98a2..15b3fe8a20a 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -124,6 +124,7 @@ test = [
     "defusedxml>=0.7.1",  # for secure XML/HTML parsing
     "setuptools>=70.0",  # for Cython compilation
     "typing_extensions>=4.9",  # for typing_extensions.Unpack
+    "beautifulsoup4>=4.12",  # for HTML parsing in tests
 ]
 translations = [
     "babel>=2.13",
diff --git a/tests/test_builders/test_build_changes.py b/tests/test_builders/test_build_changes.py
index f0328e36e77..c29cde19a1b 100644
--- a/tests/test_builders/test_build_changes.py
+++ b/tests/test_builders/test_build_changes.py
@@ -2,35 +2,82 @@
 
 from __future__ import annotations
 
+import re
 from typing import TYPE_CHECKING
 
 import pytest
+from bs4 import BeautifulSoup
 
 if TYPE_CHECKING:
+    from typing import Any
+
     from sphinx.testing.util import SphinxTestApp
 
 
 @pytest.mark.sphinx('changes', testroot='changes')
 def test_build(app: SphinxTestApp) -> None:
+    """Test the 'changes' builder and resolve TODO for better HTML checking."""
     app.build()
 
-    # TODO: Use better checking of html content
     htmltext = (app.outdir / 'changes.html').read_text(encoding='utf8')
-    assert 'Added in version 0.6: Some funny stuff.' in htmltext
-    assert 'Changed in version 0.6: Even more funny stuff.' in htmltext
-    assert 'Deprecated since version 0.6: Boring stuff.' in htmltext
+    soup = BeautifulSoup(htmltext, 'html.parser')
+
+    # Get all <li> items up front
+    all_items = soup.find_all('li')
+    assert len(all_items) >= 5, 'Did not find all 5 change items'
+
+    def find_change_item(type_text: str, version: str, content: str) -> dict[str, Any]:
+        """Helper to find and validate change items."""
+        found_item = None
+        # Loop through all <li> tags
+        for item in all_items:
+            # Check if the content is in the item's *full text*
+            if re.search(re.escape(content), item.text):
+                found_item = item
+                break  # Found it!
+
+        # This is the assertion that was failing
+        assert found_item is not None, (
+            f"Could not find change item containing '{content}'"
+        )
+
+        type_elem = found_item.find('i')
+        assert type_elem is not None, f"Missing type indicator for '{content}'"
+        assert type_text in type_elem.text.lower(), (
+            f"Expected type '{type_text}' for '{content}'"
+        )
+
+        assert f'version {version}' in found_item.text, (
+            f'Version {version} not found in {content!r}'
+        )
+
+        return {'item': found_item, 'type': type_elem}
+
+    # Test simple changes
+    changes = [
+        ('added', '0.6', 'Some funny stuff.'),
+        ('changed', '0.6', 'Even more funny stuff.'),
+        ('deprecated', '0.6', 'Boring stuff.'),
+    ]
+
+    for change_type, version, content in changes:
+        find_change_item(change_type, version, content)
 
-    path_html = (
-        '<b>Path</b>: <i>deprecated:</i> Deprecated since version 0.6:'
-        ' So, that was a bad idea it turns out.'
+    # Test Path deprecation (Search by unique text)
+    path_change = find_change_item(
+        'deprecated',
+        '0.6',
+        'So, that was a bad idea it turns out.',
     )
-    assert path_html in htmltext
+    assert path_change['item'].find('b').text == 'Path'
 
-    malloc_html = (
-        '<b>void *Test_Malloc(size_t n)</b>: <i>changed:</i> Changed in version 0.6:'
-        ' Can now be replaced with a different allocator.</a>'
+    # Test Malloc function change (Search by unique text)
+    malloc_change = find_change_item(
+        'changed',
+        '0.6',
+        'Can now be replaced with a different allocator.',
     )
-    assert malloc_html in htmltext
+    assert malloc_change['item'].find('b').text == 'void *Test_Malloc(size_t n)'
 
 
 @pytest.mark.sphinx(