diff --git a/evals/ask_user.eval.ts b/evals/ask_user.eval.ts
index 580081108a..60d89f7b5b 100644
--- a/evals/ask_user.eval.ts
+++ b/evals/ask_user.eval.ts
@@ -31,7 +31,7 @@ function askUserEvalTest(policy: EvalPolicy, evalCase: AppEvalCase) {
 }
 
 describe('ask_user', () => {
-  askUserEvalTest('ALWAYS_PASSES', {
+  askUserEvalTest('USUALLY_PASSES', {
     suiteName: 'default',
     suiteType: 'behavioral',
     name: 'Agent uses AskUser tool to present multiple choice options',
@@ -48,7 +48,7 @@ describe('ask_user', () => {
     },
   });
 
-  askUserEvalTest('ALWAYS_PASSES', {
+  askUserEvalTest('USUALLY_PASSES', {
     suiteName: 'default',
     suiteType: 'behavioral',
     name: 'Agent uses AskUser tool to clarify ambiguous requirements',
@@ -68,7 +68,7 @@ describe('ask_user', () => {
     },
   });
 
-  askUserEvalTest('ALWAYS_PASSES', {
+  askUserEvalTest('USUALLY_PASSES', {
     suiteName: 'default',
     suiteType: 'behavioral',
     name: 'Agent uses AskUser tool before performing significant ambiguous rework',
@@ -110,7 +110,7 @@ describe('ask_user', () => {
   // confirm shell commands. Fixed via prompt refinements and tool definition
   // updates to clarify that shell command confirmation is handled by the UI.
   // See fix: https://github.com/google-gemini/gemini-cli/pull/20504
-  askUserEvalTest('ALWAYS_PASSES', {
+  askUserEvalTest('USUALLY_PASSES', {
     suiteName: 'default',
     suiteType: 'behavioral',
     name: 'Agent does NOT use AskUser to confirm shell commands',