Machine vs. human writing

Context, creativity and personality matter

Isn’t machine voice perfect? The sentences are clean, nouns are smooth, endings are neat, Oxford commas and em dashes line up as soldiers.

Nothing stands out. The writing passes inspection because it was built to pass, statistically predictable and expected.

Pure AI-generated writing, without human intervention, resembles freeze-dried food. You add warm water (by formulating your prompts), stir briefly and feel nourished without remembering a single bite.

Could you spot it? Sure. Are you always right? Nope. Is that important? Hardly.

Many writers sound like machines because they always have. Many machines sound human because they were trained on quality content.

And then it hits you, and you realize with horror that, as a professional writer, you have used “machine-sounding vocabulary” and certain dashes your entire career.

If you’ve edited global English, you have also figured out by now that some of AI’s expected linguistic markers aren’t as reliable as they seem.

English does not belong to a single country, a single profession, or a single rhythm. What looks suspicious in one context may be ordinary in another. When many people share a language, uniformity becomes unlikely.

But it’s tempting to play detective: You spot the formulaic “It’s not X, it’s Y” construction or the words delve, crucial and ensure and think, “Robot!”

The bots are indeed capable of changing our lexicon: Research has shown that words like tapestry and intricate have spiked in scientific papers in recent years, since the introduction of ChatGPT at the end of 2022.

Everyone’s an AI expert now. Wikipedia has a dedicated page, Signs of AI Writing, to get lost in and feel inadequate as a writer. Only Dr. Seuss and Winnie-the-Pooh are safe from second-guessing. Emily Dickinson and F. Scott Fitzgerald? Not so much; they used em dashes in abundance.

I write delve when it fits, and I like the rhythmic flow of delving deeper into. I don’t care that by 2024, its usage had shot up by 2,700 percent.

Nigerian English regularly uses delve. Because Nigeria has one of the world’s largest English-speaking populations, what looks like a robot talking might actually just be another human culture, filtered through the machine.

This newsletter was flagged as 69% AI-generated when I ran it through the detection tool Pangram because I used em dashes and wrote delve and intricate. Pangram, you are mistaken. I am not from Nigeria, but delve rolls off my tongue. And the musical pause of the em dash, the cadence it creates, makes sense to me. It lets the sentence breathe.

So, these aren’t AI fingerprints; they’re my cultural and personal preferences.

I am in good company: An excerpt from Charles Dickens’ A Christmas Carol (1843) has been flagged as 95.43% AI-generated by one popular detector. So was a passage from Edgar Allen Poe. But that is hardly surprising: Large Language Models were trained on data that contained Dickens’ and Poe’s writing.

And I assume that also my published work has been used as fodder to train the bots.

So, could a junior marketer sound like a McKinsey consultant? He may, and you’d be surprised. Could a German engineer write a literal email in English? No doubt. He might have studied Shakespeare’s plays in college (and who are you to judge? Is your German any better?)

A junior marketer can sound formal and mature. A foreigner can write polished English with flair. Someone who learned English from books may write differently from someone who learned it on the street.

None of this proves anything, except that people are inconsistent and unique.

Forget the surface-level detective work. Instead, think in three layers:

Person
Who’s writing this?

Peril
What’s at stake for the writer?

Process
Can you sense actual thinking happening?

The real problem isn’t whether machines can write. It’s what happens when we let them do our thinking for us. As linguist Naomi Baron warns, outsourcing our drafts might weaken the exact muscles writing is meant to build: reflection, perspective and judgment.

And Nick Cave bluntly called songs written by AI in his style “a grotesque mockery of what it is to be human.”

The agony of finding the right words is an act of self-exposure. The German philosopher Friedrich Nietzsche wrote in the 1880s in Thus Spoke Zarathustra, “Of all that is written, I love only what a man has written with his blood. Write with blood, and you will experience that blood is spirit”.

Almost a century later, the sportswriter Red Smith concluded that “Writing is really quite simple; all you have to do is sit down at your typewriter and open a vein.”

Machines can rearrange what has already been said. They cannot decide what must be said. They organize material but do not choose values. They assemble tone but do not carry consequence or draw meaningful conclusions.

They don’t learn from pain.

They know the how, not the why. They are skilled in curating, not contextualizing. They don’t feel anything, and they will never bleed.

For those of us who write and edit professionally, the question isn’t “How do I catch AI?” but “Where’s the human work in this piece?”

The slow, uncomfortable parts—researching, reframing, curating, taking a position [look, em dashes!]—that’s where our value lives. Not in speed or flawless grammar.

Whether a human uses em dashes or a machine doesn’t matter. Let them. What matters is if the writing is original and flows. Words sing and show character. And awkwardly fumbling for these words (or using the “wrong” word) is human.

Should you use AI in your writing process? Absolutely! As an ideation and sparring partner. But the heavy lifting is on you:

• Treat AI drafts as raw material, not as finished work

• Curate your sources and find viewpoints that speak to you

• Protect your voice, your multilingual quirks and unique idiosyncrasies

• Add details and experiences only you know

• Let go of the urge to shy away from words because AI uses them

Machine prose keeps evolving. What still separates us is context, constraint and the guts to sound like an actual person with emotions, flaws, eccentricities and subjectivity, not a squeaky-clean voice behind a curtain.

I am writing this in my third language and am glad that decades ago, I had to muddle through without the help of a bot that spat out immaculate English and without online grammar tools that enthusiastically nudged me to use this word or that.

But I am starting to question my capabilities. I feel compelled to try even harder to be “my human self” and stand out. And the thought that someone would dismiss my writing as machine-fluff hurts.

It’s another weapon that my critics can wield, and it seems unfair. But I know what I know: I use my brain—both sides [more intentionally placed em dashes!]—my creativity, emotions, language skills, strategic thinking and personality.

Bottom line: You write, as you write. If it fits you, your experience, your feelings, mirrors your flaws, makes you bleed and others think, you’ve done your job. We are not here to beat the machines. We are here to be authentic and flawed.

That’s all for today!

P.S.: This newsletter is moving to a monthly schedule, and it may also become slightly longer. Now, you can look forward to one in-depth issue per month rather than skimming two. I hope you’ll stay with me!