Rhetoric professors in ancient Alexandria classified arguments not by their logic but by their *ethos* — the perceived character of the speaker — because they understood that persuasion travels on trust before it travels on reason. The rhetorician Quintilian spent twelve volumes on this: credibility is not a halo effect layered on top of content; it is a structural precondition for content to land at all. Now consider what happens when AI writes the communication you send. The content may be accurate, the tone perfectly calibrated — and the *ethos* is slowly becoming a performance, because the relationship between your words and your actual character is being decoupled by a tool that has no stake in the relationship. This is not a complaint about AI drafting. It's a structural observation: Quintilian's insight predicts that when receivers gradually sense a mismatch between your written voice and your spoken one, your emails, your Slack messages, your proposals — they don't just read differently. They start being unconsciously discounted.

If someone who knows you well read your last five AI-assisted messages without knowing you wrote them, what would they conclude about who you are — and how close is that to accurate?

Drawing from Classical rhetoric / Roman oratory theory — Quintilian (Institutio Oratoria, c. 95 CE)