What Makes a Story “Look” Like AI Wrote It?

I have clues!

First off, I have been primarily posting over here on Substack, but I’m going to make an effort to duplicate that content over her for those of you who are anti-Substack.

It is rare that I engage in social media threads (This is mostly because social media is scary.), but I did yesterday in a big moment of weakness.

Why?

There’s been some writing world drama about a traditionally published novelist having her book pulled after the publisher decided (post printing) that there were AI tells within it.

So, I responded to a post because it was about AI writing.

And AI writing has started to take up a lot of my time as a developmental editor.

This means I highlight sections of text and say, “This sounds like AI” a lot. So much. Sometimes it isn’t AI-helped, but a lot of times it is.

AI tells change as its development and modelling grows. It used to be all about the em-dash, which shows a break in thought—I am a mushroom—during a sentence or at its end.

That’s still a tell, so let’s get into it, okay?

Oh! And I am only talking about fiction here. There are a ton of brilliant articles about spotting AI (or making your work sound less like AI) in nonfiction and press releases and blog posts.

EM-DASH.

This doesn’t always mean that there’s been an AI assist in a novel. Human novelists have used the em-dash for ages and AI has been trained off the work of human novelists.

And these little buggers are addictive. They make you think less about clauses and commas and we lean into them to be funny or for emphasis. It’s a bit of cheat when it comes to digging deeper and working harder.

So when does the em-dash indicate AI?

Usually when:

1. There’s one in every paragraph.
2. When they aren’t actually indicating a break in thought.

Let’s zoom in on that one, okay?

If read:

Carrie was so upset with Jack—he was a hamster loving dog.

I’d think, “What the what?”

That’s not a break in thought. It’s a continuation of thought. That’s an AI tell.

That would do better as two independent sentences or with a semicolon between them. Like one of these:

Carrie was so upset with Jack; he was a hamster loving dog.

Carrie was so upset with Jack. He was a hamster loving dog.

I am now a bit intrigued by this, actually.

So, to use that em-dash appropriately to show a break in thought and not necessarily an AI tell, it would work like this:

Carrie was so upset with Jack—What was her favorite Portuguese pop song?—Peanuts are devious punks.

THE MAGICAL SENTENCE FRAGMENT PARAGRAPH.

A lot of people think that the rule of threes indicate AI, but again, good writers have been using that for emphasis for ages. What they haven’t been using is this: the broken sentence paragraphs.

In this tell, each sentence fragment is its own paragraph on the page. So like this:

But I loved him.

And I knew it.

But he had—

Hard eyes of steel Maine winters.

Those ice cold eyes.

But his smile warmed me.

Often these come at an end of a scene, right before a “hook” to read more.

It was now or never.

She opened the door.

THESE ANNOYING SENTENCES.

There is a self-declaratory feel to a lot of AI writing even when in first-person close narratives.

I’ve pulled these two examples from the NYT examples of Claude (an AI tool) that shows what I’m talking about.

Ambiguity was not weakness. It was survival.

To understand this is not to feel small. It is to feel implicated in something vast.

This is why the wise hesitate. Not from cruelty, but from understanding that interference ripples outward in ways we cannot trace. To cure a blight may curse a harvest three valleys over. Power is not the difficult thing. Restraint is the difficult thing.

These super declarative sentences with weak verbs is very AI.

THE COMMA + LIKE or + AS IF

The “comma + like” construction. AI systems often default to structures like:

• “…reflections up the trees, like a cruel joke”
• “…herded, as if livestock into slaughter”

The issue isn’t that the comparison is wrong. It’s really just that the rhythm is becoming recognizable as an AI pattern. Human prose tends to vary the phrasing a bit more, either by reordering the sentence or making the comparison more specific or more fun, honestly.

But when you have a ton of the same construction, it begins to feel like AI.

THE HUMAN ELEMENT: EMOTIONAL RESONANCE/DEEP POV/MRUs.

In fiction writing, we tend to want emotional resonance. AI does this on a sentence level often with those em-dashes, declarative sentences.

So that means it tends to avoid deep point of view and also emotional cause and effect and also motivation-reaction units.

Here’s a quick example of deeper point of view:

After a few blocks, she couldn’t walk and felt like a total fool when she realized nothing was pursuing here.

So, a deeper POV might be something like:

After a few blocks sprinting across the broken asphalt, her legs burned, and her silly lungs burned, and her whole face, too. No zombie hamsters were chasing her. Idiot. She groaned as she leaned against the flat wall of the WalMart and rubbed her thighs. What an idiot.

Oh, man. I lied. So, this isn’t just about the deepness of her point of view which is shown via that internal voice (“Idiot” “What an idiot”); it’s also about concrete detail (broken asphalt, flat wall) and the action that makes sense for the context, which also brings you deeper into the POV (sprinting, groaning, leaning, rubbing).

You can see it, right?

Motivation-reaction units are other thing that breaks down especially the longer the story is.

Dwight Swain’s Motivation-Reaction Unit (MRU) theory is basically the idea that character reactions should follow a logical order: first the stimulus (motivation), then the reaction (which itself unfolds in a logical sequence: feeling → reflex → action → speech). It’s all about logical cause and effect.

AI can often get cause and effect, but it will repeat the same cause and effect over and over again OR it will fail to have the emotion linked to it. Or it will just lose the sequence of it.

There! I could go on, but this is sort of long already and I wanted to get the major players in there. I hope it helps you spot AI, but also maybe helps remember to lean into the human part of good writing, too. Or, if you use AI to write your stories, helps you know what to look for to make your story stronger.

You’ve got this, okay?

REEDSY STUDIO THINGS

So, there’s some cool resources on Reedsy for building a character, including a character template. I think I might get two cents or something if you go through it via the link because I’m now an affiliate, which means I get to say, “I may receive a commission if someone makes a purchase through your affiliate link.” But mostly there are some cool tools on Reedsy that you might want to check out. No pressure though!