You can also write using sound based/compressed 'text message' dialect: unless a real human is reading, automated watching tool should have a hard time (until coded/ML-ed on such dialects I guess)