Why Does AI Write Like That?

|

,

In a previous post, I mentioned reading The New Yorker article “What Kind of Writer Is ChatGPT?” by Carl Newton. The piece profiles Chris, a graduate student who used ChatGPT to generate a draft paragraph for a writing assignment.

Continuing that line of inquiry, I recently read a New York Times Magazine essay, “Why Does A.I. Write Like … That?” by Sam Kriss, where he examines the unsettling rise of A.I.-generated prose, and the growing sense that something is quietly going wrong with how the world now speaks.

Kriss highlights that A.I. writing is everywhere. It fills news articles, emails, student essays, corporate statements, social media posts, and even restaurant menus. Its growing presence raises an uncomfortable question: What happens to creativity, authenticity, and human voice when machines produce so much of our language?

Kriss finds A.I. writing deeply unsettling. It sounds familiar, more polished, confident, competent, but also hollow. It gestures toward meaning without quite landing it. The words are smooth, but the experience is strangely empty.

He describes a hard-to-pin-down feeling many readers recognize: A.I. prose feels omnipresent, and once you notice it, it’s hard to unhear. Kriss explores how this voice took hold, how to spot it, and why it matters, for creativity, culture, and what it means to sound human.

If you’re anything like me, you did not enjoy reading that paragraph. Everything about it puts me on alert: Something is wrong here; this text is not what it says it is. It’s one of them.

Entirely ordinary words, like “tapestry,” which has been innocently describing a kind of vertical carpet for more than 500 years, make me suddenly tense. I’m driven to the point of fury by any sentence following the pattern “It’s not X, it’s Y.”

But whatever these quirks of language used to mean, that’s not what they mean anymore. They’re now telltale signs that what you’re reading was churned out by an A.I.

– Sam Kriss, The New York Times

Below are the concisely summarized key highlights of Kriss article:

  • A voice that’s everywhere: A.I.-generated writing is no longer a novelty. One uncredited “author” now produces a staggering share of the world’s text. This isn’t just a content overload, it’s a shift in how language itself sounds.
  • Why A.I. writing feels “off”: Kriss notes that at first glance, A.I. prose appears elegant and capable. But many readers experience a vague unease. The writing lacks lived experience. Sensory details feel abstract, emotions are inflated, and metaphors pile up without grounding. The result is language that’s technically correct but emotionally hollow.
  • A style with telltale tics: Despite claims of neutrality, A.I. writing has a very recognizable voice. It loves em dashes. It leans on formulas like “It’s not X, it’s Y” or “No X. No Y. Just Z.” It overuses words like delve, tapestry, quiet, echo, and journey. These habits come from mistaking the surface features of good writing for the thing itself.
  • From weird and funny to smooth and boring: Early A.I. models were unpredictable and often unintentionally hilarious. Their mistakes produced genuine surprise. But with the arrival of ChatGPT, the chaos was smoothed out almost overnight. Today, A.I. is mostly used for practical tasks, emails, summaries, public statements, efficient, fluent, and deeply forgettable.
  • When A.I. copies culture and spreads it: Because A.I. is trained on vast amounts of online text, it absorbs regional language patterns and redistributes them elsewhere. Words like delve surged partly due to Nigerian English usage. Phrases such as “I rise to speak” have appeared in British Parliament, likely imported from American political speech. Culture doesn’t just pass through the machine, it leaks.
  • The fiction problem: ghosts, whispers, and quiet: Ask A.I. to write fiction and patterns emerge quickly. Everything is quiet. Everything hums. Characters become shadows, echoes, and ghosts. A.I. knows good writing values subtlety, so it exaggerates subtlety until it becomes meaningless—confusing the appearance of depth with depth itself.
  • Language without experience: A.I. cannot see, taste, touch, or remember. Its sensory language floats free of reality. Grief tastes like metal. Rooms smell like dreams. Meaning collapses under abstraction. The writing sounds profound but often says very little.
  • The real risk: we start sounding like it: Kriss argues that the greatest danger isn’t that A.I. writes for us, but that we begin to write like A.I. Studies already show humans picking up its rhythms and habits. Many people find this language emotional, beautiful, even comforting. Over time, the machine’s voice may quietly become our own.

Finally, Kriss emphasizes that the real threat of A.I. writing isn’t replacement but it’s influence. As we absorb more machine-generated language, we risk losing the texture, specificity, and messiness that come from human experience. What’s at stake isn’t just how we write, but how we think and how we sound when we’re most ourselves.

When I asked ChatGPT whether it agreed with Kriss’s view, this is how it responded:

I largely agree with Kriss. A.I. writing tends to produce a polished but flattened voice—emotionally correct yet experientially thin—because it’s built from patterns, not lived experience. The real risk isn’t that A.I. replaces human writing, but that people unconsciously adopt its rhythms, abstractions, and clichés. That said, this outcome isn’t inevitable. Used thoughtfully, A.I. can sharpen the human voice rather than erase it. The danger lies not in the tool itself, but in forgetting to notice how we sound when we use it.

Acknowledgement: This summary was generated using A.I. writing tools as part of an experimental learning process. For deeper context, please refer to the original essay in The New York Times.

Related resource Links