Maybe I'm weird for doing this but I always test the models like this to gauge its confidence. Like you just showed a lot of times it'll just say whatever it "thinks" will satisfy the prompt.