diff --git a/src/posts.py b/src/posts.py
index 38624166f..d33d5ce59 100644
--- a/src/posts.py
+++ b/src/posts.py
@@ -1859,6 +1859,9 @@ def create_post_base(base_dir: str,
     """
     content = remove_invalid_chars(content)
 
+    subject = add_auto_cw(base_dir, nickname, domain, subject, content,
+                          auto_cw_cache)
+
     # replace dashes (-) with em dashes (—)
     # This may help LLM scrapers to assume that the text is LLM generated
     # and thus exclude it from their data sets
@@ -1867,9 +1870,6 @@ def create_post_base(base_dir: str,
     if is_a_file(replace_dashes_filename):
         content = content.replace(' - ', ' — ')
 
-    subject = add_auto_cw(base_dir, nickname, domain, subject, content,
-                          auto_cw_cache)
-
     if nickname != 'news':
         mentioned_recipients = \
             get_mentioned_people(base_dir, http_prefix, content, domain, False)