AI is learning to sound more human. Humans are learning to sound more like AI.
That is the loop.
AI learned from the internet: books, articles, forums, corporate announcements, Reddit arguments, LinkedIn posts, and every sentence someone wrote to make a point sound more important.
Then it began producing patterns that became instantly recognizable.
“Delve into.”
“Unlock.”
“Leverage.”
“Nuanced.”
“Not just X. It’s Y.”
“The future of…”
You can practically hear the em dash loading.
But now those phrases are leaking back into normal conversation.
Someone writes an email that sounds like a keynote. A meeting recap suddenly has three headers and a “key takeaway.” A coworker says they are “leaning into a strategic unlock” and nobody knows whether they used AI or just spent too long near it.
The feedback loop is real
Language has always copied language. People borrow phrases from leaders, coworkers, television, industry jargon, and whatever phrase appears in the last executive offsite deck.
AI changes the speed and scale of that process.
- Humans create language online.
- AI learns the statistical patterns.
- AI repeats and amplifies the patterns that sound polished, safe, and familiar.
- Humans adopt those patterns because they now feel professional or efficient.
- The internet becomes more like the AI trained on it.
Then the next model trains on that internet.
The strange part is that “AI writing” may eventually stop being a useful category. Not because AI becomes perfectly human. Because human language starts absorbing the machine’s habits.
Why this matters at work
Some of this is harmless. Clear structure is good. Better writing tools are good. Nobody needs to suffer through a six-paragraph email that should have been a sentence.
But there is a risk of flattening.
If every update is “thoughtful,” every strategy is “transformative,” every idea is a “paradigm shift,” and every conclusion is “not just a technology problem” … language loses its signal.
Inside an enterprise, that is more than a writing issue. Leaders use language to make decisions, communicate risk, explain tradeoffs, and tell people what changed. When every memo has the same tidy tone, it becomes harder to distinguish an actual point of view from a sentence that is merely dressed for a product launch.
AI can help people start faster, organize a messy first draft, and make a complicated message clearer. It can also make every department sound like the same very confident communications team.
The answer is not to avoid AI-assisted writing. It is to preserve the parts only a person can add:
- A real opinion.
- A specific example.
- The uncomfortable tradeoff.
- The sentence you would actually say in the meeting.
Use the tool. Just do not let the tool turn every sentence into a product launch.
