Attacking bot comments based on communication style doesn't hurt AI's feelings, but it does act as a red flag for humans who communicate differently