AI responses are getting more harder to detect, down the road we wouldn't take which is genuine and which isn't