Everyone Knows You Didn’t Write That

How AI-polished messages quietly break trust, and when using AI is still the right call.

The patch holds. The hole is still there.

Many of us have seen, at some point in the last two years, a message like this:

Dear Team, I hope this message finds you well. First and foremost, I want to acknowledge the incredible dedication each and every one of you has demonstrated over the past quarter. Your commitment does not go unnoticed — and I am truly grateful to be part of such a talented group. Regarding the PIM Launch After careful consideration and alignment with key stakeholders, we have made the decision to adjust the timeline. While I understand this may raise questions, I want to assure you that this decision was not made lightly — it reflects our ongoing commitment to delivering excellence rather than rushing outcomes. Next Steps We will be sharing more details in the coming weeks. As always, my door is open — please don’t hesitate to reach out with any questions or concerns. Best regards, X (someone high up the ladder)

The overly formal style. The robotic lack of personality. The em dashes. The fog in the content. Nobody truly reads past the second paragraph. I have seen messages like this more and more often, sometimes delivered, awkwardly, as a Slack message in the format of a letter. They tend to come from the top: C-level, HR, directors.

And yes, I used AI while editing this article. I will come back to why that is not a contradiction.

Most people use AI in written communication to gain clarity and professionalism. But there are other reasons underneath, conscious or not. Lack of confidence. Fear of how words written plainly by hand will land, or how they could be used against us. That fear pushes us to run every message through a filter.

Sometimes this is not a bad thing. We are emotional. Using AI to strip sarcasm, anger or disappointment out of a message can genuinely improve collaboration. Certainly helped me with my ex and co-parenting!

The problem starts when AI replaces the thinking instead of the emotion. And it leaks. You can see it in the gap between how a leader communicates in person and in writing. In the meeting: shallow answers, no solid logic, ambiguity, impatience. An hour later on Slack: refined, polite, understanding. The contrast is what does the damage. The reader’s reaction is almost a reflex: Is he even listening? What is he hiding? Why doesn’t he write like himself?

A team I know spent several sessions preparing a product strategy document for review. The feedback from above arrived as a single block of pasted text: numbered points, balanced tone, impeccable structure. It landed as “that’s my feedback, handle it.” Nobody could find his vocabulary anywhere in it. The team spent more time debating whether he had actually read the document than discussing the points themselves.

That is what detection costs. The hierarchy keeps such leaders in charge. It does not make people trust them or follow them.

“Writing is nature’s way of letting you know how sloppy your thinking is,” said the computer scientist Leslie Lamport. When leaders outsource their thinking to AI, in messages, emails, LinkedIn posts, the words exist but the thought behind them does not. Context becomes secondary. Generic bad ideas get implemented. And when credibility in leadership breaks down, the foundations that make an organisation work start to slowly crumble.

So why do we do it? Two reasons: we want to look good, and we are afraid.

Social life is a performance, with a front stage and a back stage [1]. AI gives everyone a professional ghostwriter for the front stage. The leader’s problem is that his back stage, the meetings, the real-time questions, is now publicly comparable with the front.

There is a catch, though. People value the same “artifact” more when they believe more effort went into it [2]. Think of handmade leather shoes versus the factory pair. Same function, different worth. A written message used to work the same way: it carried proof that someone spent time thinking about you. AI removed the cost, and with it, the proof.

Fear is the other reason. In a low-trust environment, the sanitised message is a rational defence. I know the feeling from the receiving end. I once needed a straight answer from someone above me on a delicate situation. I asked directly, in writing. The replies came slow, and in a voice that was not his: careful, balanced, giving the shape of an answer without the substance. Somewhere in that exchange a question crossed my mind that I could not put back: is he pasting my messages into ChatGPT? I have no proof that he was and that is exactly the point. Once the question exists, you read every following reply through it. I noticed I had started writing my own messages more defensively too. Two people filtering themselves through a machine, in a conversation whose only purpose was honesty.

None of this makes AI hopeless for communication. When your ideas and your voice stay attached to the message, AI is an amplifier: clarity, language range, an intelligent autocorrect, an emotional filter that keeps conversations objective.

English is my second language. When I write something delicate, I draft it myself first, then ask AI what a native speaker might misread or where my tone lands harder than I intended. That is how this article was written too: the way you would work with a human editor. It fixed my grammar, questioned my structure, and occasionally saved me from a sentence that sounded like an accusation when I meant a question. The writing reflects the books I’ve read and the ideas that have influenced me. But my contribution lies in how I understand them, connect them, and decide what I want to say. The structure, the examples, and the choice of what to include or leave out are mine. For me, that’s the line to hold.

The next time you send a message with AI, check one thing: can you find yourself in those words?

The message was perfect. That’s how everyone knew you weren’t in it.

Further reading

[1] Goffman, The Presentation of Self in Everyday Life (1959) [2] Kruger et al., “The Effort Heuristic,” Journal of Experimental Social Psychology (2004)

• Jakesch et al., “AI-Mediated Communication” CHI (2019), on how suspicion of AI authorship reduces trust. • Hohenstein et al., Scientific Reports (2023), on AI-mediated messages and relationship perception.