diff --git a/cursorless-talon/src/get_grapheme_spoken_form_entries.py b/cursorless-talon/src/get_grapheme_spoken_form_entries.py
index d50fb0d98e..42ff196b00 100644
--- a/cursorless-talon/src/get_grapheme_spoken_form_entries.py
+++ b/cursorless-talon/src/get_grapheme_spoken_form_entries.py
@@ -14,16 +14,6 @@
 def get_grapheme_spoken_form_entries(
     grapheme_talon_list: dict[str, str],
 ) -> list[SpokenFormOutputEntry]:
-    if grapheme_capture_name not in registry.captures:
-        # We require this capture, and expect it to be defined. We want to show a user friendly error if it isn't present (usually indicating a problem with their community.git setup) and we think the user is going to use Cursorless.
-        # However, sometimes users use different dictation engines (Vosk, Webspeech) with entirely different/smaller grammars that don't have the capture, and this code will run then, and falsely error. We don't want to show an error in that case because they don't plan to actually use Cursorless.
-        if "en" in scope.get("language", {}):
-            app.notify(f"Capture <{grapheme_capture_name}> isn't defined")
-            print(
-                f"Capture <{grapheme_capture_name}> isn't defined, which is required by Cursorless. Please check your community setup"
-            )
-        return []
-
     return [
         {
             "type": "grapheme",
@@ -37,6 +27,16 @@ def get_grapheme_spoken_form_entries(
 
 
 def get_graphemes_talon_list() -> dict[str, str]:
+    if grapheme_capture_name not in registry.captures:
+        # We require this capture, and expect it to be defined. We want to show a user friendly error if it isn't present (usually indicating a problem with their community.git setup) and we think the user is going to use Cursorless.
+        # However, sometimes users use different dictation engines (Vosk, Webspeech) with entirely different/smaller grammars that don't have the capture, and this code will run then, and falsely error. We don't want to show an error in that case because they don't plan to actually use Cursorless.
+        if "en" in scope.get("language", {}):
+            app.notify(f"Capture <{grapheme_capture_name}> isn't defined")
+            print(
+                f"Capture <{grapheme_capture_name}> isn't defined, which is required by Cursorless. Please check your community setup"
+            )
+        return {}
+
     return {
         spoken_form: id
         for symbol_list in generate_lists_from_capture(grapheme_capture_name)
diff --git a/cursorless-talon/src/spoken_forms.py b/cursorless-talon/src/spoken_forms.py
index 1dbb3b3efd..1b89cd96c5 100644
--- a/cursorless-talon/src/spoken_forms.py
+++ b/cursorless-talon/src/spoken_forms.py
@@ -125,6 +125,7 @@ def handle_new_values(csv_name: str, values: list[SpokenFormEntry]):
         if initialized:
             # On first run, we just do one update at the end, so we suppress
             # writing until we get there
+            init_scope_spoken_forms(graphemes_talon_list)
             update_spoken_forms_output()
 
     handle_csv = auto_construct_defaults(
diff --git a/cursorless-talon/src/spoken_scope_forms.py b/cursorless-talon/src/spoken_scope_forms.py
index ff6e8b07e2..da72d85a1e 100644
--- a/cursorless-talon/src/spoken_scope_forms.py
+++ b/cursorless-talon/src/spoken_scope_forms.py
@@ -4,25 +4,20 @@
 
 
 def init_scope_spoken_forms(graphemes_talon_list: dict[str, str]):
-    create_flattened_talon_list(
-        csv_get_ctx(), graphemes_talon_list, include_custom_regex=True
-    )
+    create_flattened_talon_list(csv_get_ctx(), graphemes_talon_list)
     if is_cursorless_test_mode():
-        create_flattened_talon_list(
-            csv_get_normalized_ctx(), graphemes_talon_list, include_custom_regex=False
-        )
+        create_flattened_talon_list(csv_get_normalized_ctx(), graphemes_talon_list)
 
 
-def create_flattened_talon_list(
-    ctx: Context, graphemes_talon_list: dict[str, str], include_custom_regex: bool
-):
+def create_flattened_talon_list(ctx: Context, graphemes_talon_list: dict[str, str]):
     lists_to_merge = {
         "cursorless_scope_type": "simple",
         "cursorless_surrounding_pair_scope_type": "surroundingPair",
         "cursorless_selectable_only_paired_delimiter": "surroundingPair",
         "cursorless_wrapper_selectable_paired_delimiter": "surroundingPair",
     }
-    if include_custom_regex:
+    # If the user have no custom regex scope type, then that list is missing from the context
+    if "user.cursorless_custom_regex_scope_type" in ctx.lists.keys():  # noqa: SIM118
         lists_to_merge["cursorless_custom_regex_scope_type"] = "customRegex"
 
     scope_types_singular: dict[str, str] = {}