Why your writing sounds like AI (and what to do about it)

At some point recently, you read something at work and thought — this was written by AI. And then consciously or not, you disengaged. You may not have been able to name what gave it away, but you stopped reading closely.

If AI is going to be useful in moments that matter most, the writing it produces can’t sound like it came from AI. Understanding why it sounds off is the first step to fixing it.

AI-sounding writing is usually overly balanced, carefully polite, logically complete, and emotionally weightless. It explains everything and commits to very little. That’s because AI is optimized to remove friction. It smooths edges, fills gaps, and resolves tension, even when tension is exactly what a human reader expects to feel.

Some people try to fix this by adding typos or small signals to suggest a human was behind the keyboard. But making the content less polished doesn’t address what readers are actually responding to.

Human writing, especially at work, usually contains uneven emphasis, selective omission, an implicit point of view, and a sense of stake. When those are missing, the writing doesn’t feel authentic.

This is why people still send emails that sound like AI and thus fall flat. They accepted AI’s first pass as finished, even though first passes are where voice is thinnest. Once you start noticing it in others’ writing, you’ll see it everywhere.

The fix isn’t to stop using AI to draft. The mistake is stopping there.

AI is good at producing a reasonable version of something. Your voice comes from what you decide to change afterward. You can avoid sounding like AI by choosing.

After AI gives you a draft, ask yourself: What is the one point I actually care about? Where do I want the reader to pause, not skim? Make a choice about what matters most, then cut or downplay everything else.

In practice, AI generated content tends to be heavy on exposition. It earns its way into ideas slowly, which means I often cut the first paragraph entirely. A more compelling opening is usually already there, just buried.

Watch for the places where AI gives you three examples or three variations of the same point. That’s not depth — it’s hedging. Ask whether each one is doing distinct work or whether one of them is simply the best. Pick that one and cut the rest.

The same logic applies to steps and recommendations. AI will often give you five when you only need two. The ones worth keeping are the ones you actually have something to say about. Go deeper on those and let the others go.

Each of these edits is a judgment call and reasserts your voice.

For most communications at work, that’s enough because sounding a little generic is usually harmless.

There are times, though, when your audience needs to feel that you are there with them — responding to their situation, not to a situation in general. Delivering disappointing news. Responding to concern or frustration. Acknowledging effort or loss. Taking responsibility. Reassuring someone who feels exposed.

AI-polished language can feel distancing, even when it says the right things. What gives it away is lack of specificity. Humans reference the actual situation, the particular person, the real impact. AI defaults to general empathy.

For moments like these, AI is still useful — just not for drafting the final message. Use it to think through what the person might be feeling, identify what needs to be acknowledged, and pressure-test what not to say. Then write the message yourself.

Imperfect wording with real specificity builds more trust than flawless phrasing that could apply to anyone.

Knowing when not to let AI speak is as important as knowing how to use it well.