You’ve probably caught yourself doing it. You type an email, and the word “delve” slips out. You write “it’s important to note that” when you used to just say “look.” You hedge, you soften, you structure your thoughts in tidy bullet points with a neat little summary at the end.
You didn’t talk like this two years ago. Nobody did.
The tool you use to sound smarter is quietly making you sound like everyone else.
Here’s what’s actually happening, and why it should unsettle you more than any sci-fi scenario about rogue AI: we built language models by feeding them billions of words of human writing. Now those models are feeding their words back to us. We adopt their phrasing, their rhythm, their cautious hedging. We write something with AI assistance, publish it, and that output becomes training data for the next model. It’s a feedback loop with no off switch.
Most people notice the vocabulary. “Delve.” “Tapestry.” “Navigate.” “Robust.” The AI tells are easy to mock. But that’s surface noise. The real shift is structural, and it’s cognitive.
Think about how ChatGPT constructs an argument. It hedges. It qualifies. It presents multiple perspectives with diplomatic even-handedness before arriving at a careful, non-threatening conclusion. It apologizes. It says things like “while it’s true that X, it’s also worth considering Y.”
Now listen to yourself in meetings.
You’re not just borrowing AI’s words. You’re borrowing its worldview — one where every opinion comes wrapped in padding and every conviction arrives pre-diluted.
This is the paradox nobody wants to confront. We turned to AI to enhance our communication, to make us clearer, sharper, more persuasive. Instead, we’re unconsciously mimicking a machine that was designed to be helpful, harmless, and aggressively inoffensive. The result is a slow homogenization of human expression. Not just vocabulary. Not just style. The actual texture of how we reason.
Linguistic diversity isn’t some academic concern. It’s the raw material of thought itself. When everyone starts constructing sentences the same way — same cadence, same hedging, same careful balancing act before every claim — we don’t just sound identical. We start thinking in narrower grooves. The idiosyncratic phrase, the blunt opinion, the messy human tangent that leads somewhere unexpected — those become endangered species.
Adam Aleksic, who studies exactly this phenomenon, points out that technology has always shaped language. The printing press standardized spelling. Texting gave us abbreviations. But those were tools we controlled. This is different. AI doesn’t just give us new words. It gives us new patterns of thought and then rewards us for using them, because AI-assisted writing gets polished, approved, published, and amplified.
For the first time in history, the machine isn’t just reflecting how we speak. It’s teaching us how to speak back.
And here’s the part that should genuinely worry you: you probably can’t tell anymore which of your own thoughts are yours. That email you wrote last week — the one with the perfect structure and the diplomatic phrasing — was that you, or was that the ghost of a model trained on a million corporate memos? When you softened your criticism because “it felt more appropriate,” was that your judgment or the model’s preference for harmlessness bleeding into your instincts?
The line between human expression and machine output isn’t just blurring. For a lot of people, it’s already gone, and they don’t even know it.
So here’s the uncomfortable truth: if you use AI tools regularly — and at this point, who doesn’t — you are part of the loop. Your voice is shifting. Your argumentation is softening. Your spontaneity is being sanded down into something smoother, safer, and more uniform. Not because you chose that, but because the tool you reach for a hundred times a day was optimized for a kind of pleasant, structured, controversy-averse communication that’s now leaking into everything you write.
Reclaiming your voice isn’t about quitting AI. It’s about noticing the theft in real time — and choosing, deliberately, to sound like a human even when the machine offers you something tidier.
The next time you catch yourself writing “it’s worth noting that,” stop. Ask: would I actually say this out loud? If the answer is no, delete it. Say the blunt thing. Take the tangent. Leave the mess in.
Your weird, imperfect, unmistakably human voice is the one thing AI can’t generate. Don’t let it train that out of you too.
FAQ
Q: Isn't this just the same panic people had about texting ruining language?
A: No. Texting gave us new shorthand; it didn't restructure how we build arguments. AI doesn't just add vocabulary — it models a specific style of hedging, balancing, and diluting conviction that leaks into how we reason, not just how we type.
Q: So what am I supposed to do, stop using AI entirely?
A: That's not realistic and not the point. The practical move is awareness: notice when AI phrasing replaces your natural voice. If you wouldn't say it out loud, don't write it. Use AI for research and drafts, but the final voice should be unmistakably yours.
Q: Isn't homogenized, polite communication actually a good thing?
A: Only if you think uniform, hedging, controversy-averse expression is desirable. Linguistic diversity isn't aesthetic vanity — it's the substrate of original thought. When everyone reasons in the same structure, you get consensus without creativity. That's not progress. That's entropy disguised as polish.