diff --git a/CHANGES b/CHANGES
index e7099805500..72db33f7b8a 100644
--- a/CHANGES
+++ b/CHANGES
@@ -75,6 +75,9 @@ Bugs fixed
 
 * #11079: LaTeX: figures with align attribute may disappear and strangely impact
   following lists
+* #11093: LaTeX: fix "multiply-defined references" PDF build warnings when one or
+  more reST labels directly preceed an :rst:dir:`py:module` or :rst:dir:`automodule`
+  directive. Patch by Bénédikt Tran (picnixz)
 * #11110: LaTeX: Figures go missing from latex pdf if their files have the same
   base name and they use a post transform.  Patch by aaron-cooper
 * LaTeX: fix potential color leak from shadow to border of rounded boxes, if
diff --git a/sphinx/writers/latex.py b/sphinx/writers/latex.py
index 684164a1f7d..efb84735398 100644
--- a/sphinx/writers/latex.py
+++ b/sphinx/writers/latex.py
@@ -1504,14 +1504,14 @@ def add_target(id: str) -> None:
         if node.get('ismod', False):
             # Detect if the previous nodes are label targets. If so, remove
             # the refid thereof from node['ids'] to avoid duplicated ids.
-            def has_dup_label(sib: Element | None) -> bool:
+            def has_dup_label(sib: Node | None) -> bool:
                 return isinstance(sib, nodes.target) and sib.get('refid') in node['ids']
 
-            prev: Element | None = get_prev_node(node)
+            prev = get_prev_node(node)
             if has_dup_label(prev):
                 ids = node['ids'][:]  # copy to avoid side-effects
                 while has_dup_label(prev):
-                    ids.remove(prev['refid'])
+                    ids.remove(prev['refid'])  # type: ignore
                     prev = get_prev_node(prev)
             else:
                 ids = iter(node['ids'])  # read-only iterator
diff --git a/tests/test_build_latex.py b/tests/test_build_latex.py
index de45f5f0c5e..faf1a47b70b 100644
--- a/tests/test_build_latex.py
+++ b/tests/test_build_latex.py
@@ -1720,18 +1720,21 @@ def count_label(name):
         return content.count(text)
 
     pattern = r'\\phantomsection\\label\{\\detokenize\{index:label-(?:auto-)?\d+[a-z]*}}'
-    expect_labels = {match.group() for match in re.finditer(pattern, content)}
-    result_labels = set()
+    # labels found in the TeX output
+    output_labels = frozenset(match.group() for match in re.finditer(pattern, content))
+    # labels that have been tested and occurring exactly once in the output
+    tested_labels = set()
 
     # iterate over the (explicit) labels in the corresponding index.rst
-    for rst_label_name in {
+    for rst_label_name in [
         'label_1a', 'label_1b', 'label_2', 'label_3',
         'label_auto_1a', 'label_auto_1b', 'label_auto_2', 'label_auto_3',
-    }:
+    ]:
         tex_label_name = 'index:' + rst_label_name.replace('_', '-')
         tex_label_code = r'\phantomsection\label{\detokenize{%s}}' % tex_label_name
         assert content.count(tex_label_code) == 1, f'duplicated label: {tex_label_name!r}'
-        result_labels.add(tex_label_code)
+        tested_labels.add(tex_label_code)
 
-    # sort the labels for a better visual diff, if any
-    assert sorted(result_labels) == sorted(expect_labels)
+    # ensure that we did not forget any label to check
+    # and if so, report them nicely in case of failure
+    assert sorted(tested_labels) == sorted(output_labels)