Machines Learned It from Humans

Writers may have more than deadlines to worry about. Apparently, em dashes now belong to AI-generated writing. Polished prose. Fragments, too. It’s like saying a good essay can only be written with AI assistance, if not entirely generated by it.

Instead of getting credit for good writing, human writers now must prove they wrote their own polished prose. Built a balanced structure. Developed a compelling plot.

It’s outrageous that I even now must qualify writers as humans.

Weren’t we just writers?

I wonder what’s next. Human nurses? Human doctors? Human parents, perhaps?

I’m not just being paranoid about getting accused of using AI for even a single sentence. There’s a reason for it.

Actually, it got worse after my editor-friend Mar-Vic Cagurangan posted something on social media.

She reposted a BBC article about Zimbabwean author Simbarashe Kundizeza denying allegations that he used AI to write his Caine Prize-winning story.

The Atlantic, on Oct. 11, 2026, claimed that the AI-detection tool Pangram flagged Kundizeza’s “The Blue Fighter” as almost entirely machine-generated. Recurring phrase patterns such as sentence structures and frequent metaphors are apparent indicators of AI-generated prose.

Manoy Vallejera commented on Mar-Vic’s post: “I also scanned a story that I wrote organically and the result said it was AI. LOL! As you say, Large Language Models learned by scanning existing stories and now claim to have the authority to pass judgment on writing. SMH.”

My curiosity piqued, I pasted an essay I wrote in 2010 into ChatGPT and asked it to check if it thought the piece was AI-generated.

ChatGPT was gentler in passing judgment. It said my essay “reads as though it could be AI-assisted, especially in the second half—but the text alone cannot establish who wrote it.”

“A human can write this way, too.”

The heck! A human can write this way too? Really?!

Shouldn’t it be the other way around?

Machines can write this way too.

What created the impression that my 2010 essay was AI-assisted, according to ChatGPT, was the “accumulation of familiar rhetorical patterns.”

What gives? Apparently, I use too many metaphors, polish my phrases in batches, keep hammering home the lesson, and can’t resist tacking on extra endings. That’s what the writing machine says, anyway.

Yes. I called it.

If we’re being forced to qualify ourselves as “human” writers, AI rightfully deserves to be called what it is.

Not a machine writer. No.

A machine that writes.

It can write. But only because humans taught it how to.

The damn machine learned it from humans.

But here’s the sad part. It’s not just AI judging us now.

In the case of Kundizeza, another human writer – journalist Vauhini Vara – made the allegation, citing recurring sentence patterns and an abundance of metaphors that she claimed resemble AI-generated prose, “as well as stylistic differences” compared to Kundizeza’s “The Whistleblower” story in 2021.

Funny. Those are the same things ChatGPT identified as creating the impression that my 2010 essay was “possibly AI-assisted.”

But AI doesn’t hold exclusive rights to any stylistic patterns.

Repetition, metaphor, parallel phrasing, emotional crescendos can appear in AI output.

They also belong to human writing.

In fact, they belonged to human writers long before machines claimed pseudo-ownership.