diff --git a/swh/indexer/storage/__init__.py b/swh/indexer/storage/__init__.py
--- a/swh/indexer/storage/__init__.py
+++ b/swh/indexer/storage/__init__.py
@@ -215,6 +215,7 @@
                 default)
 
         """
+        self._check_duplicates(mimetypes, 'id')
         db.mktemp_content_mimetype(cur)
         db.copy_to(mimetypes, 'tmp_content_mimetype',
                    ['id', 'mimetype', 'encoding', 'indexer_configuration_id'],
@@ -299,6 +300,7 @@
                 default)
 
         """
+        self._check_duplicates(languages, 'id')
         db.mktemp_content_language(cur)
         # empty language is mapped to 'unknown'
         db.copy_to(
@@ -368,6 +370,8 @@
                   line, lang
 
         """
+        self._check_duplicates(ctags, 'id')
+
         def _convert_ctags(__ctags):
             """Convert ctags dict to list of ctags.
 
@@ -448,7 +452,7 @@
             list: content_license entries which failed due to unknown licenses
 
         """
-        # Then, we add the correct ones
+        self._check_duplicates(licenses, 'id')
         db.mktemp_content_fossology_license(cur)
         db.copy_to(
             ({
@@ -546,6 +550,8 @@
                 or skip duplicates (false, the default)
 
         """
+        self._check_duplicates(metadata, 'id')
+
         db.mktemp_content_metadata(cur)
 
         db.copy_to(metadata, 'tmp_content_metadata',
@@ -613,6 +619,8 @@
               or skip duplicates (false, the default)
 
         """
+        self._check_duplicates(metadata, 'id')
+
         db.mktemp_revision_metadata(cur)
 
         db.copy_to(metadata, 'tmp_revision_metadata',
@@ -665,6 +673,8 @@
               or skip duplicates (false, the default)
 
         """
+        self._check_duplicates(metadata, 'origin_id')
+
         db.mktemp_origin_intrinsic_metadata(cur)
 
         db.copy_to(metadata, 'tmp_origin_intrinsic_metadata',
@@ -754,3 +764,19 @@
         if not idx:
             return None
         return dict(zip(db.indexer_configuration_cols, idx))
+
+    def _check_duplicates(self, data, column):
+        """
+        If any two dictionaries in `data` have the same value for the
+        column, raises a `ValueError`.
+
+        Prevents 'psycopg2.ProgrammingError: ON CONFLICT DO UPDATE '
+        'command cannot affect row a second time' when conflict_update=True.
+
+        Args:
+            data (List[dict]): List of dictionaries to be inserted
+            column (str): Name of the column that acts as id.
+        """
+        if len({item[column] for item in data}) < len(data):
+            raise ValueError(
+                'The same {} is present more than once.'.format(column))
diff --git a/swh/indexer/storage/in_memory.py b/swh/indexer/storage/in_memory.py
--- a/swh/indexer/storage/in_memory.py
+++ b/swh/indexer/storage/in_memory.py
@@ -128,6 +128,10 @@
               (true) or skip duplicates (false)
 
         """
+        data = list(data)
+        if len({x['id'] for x in data}) < len(data):
+            # For "exception-compatibility" with the pgsql backend
+            raise ValueError('The same id is present more than once.')
         for item in data:
             item = item.copy()
             tool_id = item.pop('indexer_configuration_id')
diff --git a/swh/indexer/tests/storage/test_storage.py b/swh/indexer/tests/storage/test_storage.py
--- a/swh/indexer/tests/storage/test_storage.py
+++ b/swh/indexer/tests/storage/test_storage.py
@@ -1399,6 +1399,50 @@
         # metadata did change as the v2 was used to overwrite v1
         self.assertEqual(actual_metadata, expected_metadata_v2)
 
+    def test_revision_metadata_add_duplicate_twice(self):
+        # given
+        tool_id = self.tools['swh-metadata-detector']['id']
+
+        metadata_rev1 = {
+            'id': self.revision_id_2,
+            'translated_metadata': {
+                'identifier': {'@id': 'foo'},
+            },
+            'mappings': ['mapping1', 'mapping2'],
+            'indexer_configuration_id': tool_id
+        }
+
+        metadata_rev2 = {
+            'id': self.revision_id_2,
+            'translated_metadata': {
+                'identifier': {'@id': 'bar'},
+            },
+            'mappings': ['mapping1', 'mapping2'],
+            'indexer_configuration_id': tool_id
+        }
+
+        # when
+        self.storage.revision_metadata_add([metadata_rev1])
+
+        with self.assertRaises(ValueError):
+            self.storage.revision_metadata_add(
+                [metadata_rev2, metadata_rev2],
+                conflict_update=True)
+
+        # then
+        actual_metadata = list(self.storage.revision_metadata_get(
+            [self.revision_id_2, self.revision_id_1]))
+
+        expected_metadata = [{
+            'id': self.revision_id_2,
+            'translated_metadata': {
+                'identifier': {'@id': 'foo'},
+            },
+            'mappings': ['mapping1', 'mapping2'],
+            'tool': self.tools['swh-metadata-detector']
+        }]
+        self.assertEqual(actual_metadata, expected_metadata)
+
     def test_origin_intrinsic_metadata_get(self):
         # given
         tool_id = self.tools['swh-metadata-detector']['id']