Q: What is AI-voice and--honestly?--what's the harm?
"What is AI voice, exactly?"
Welcome to our weekend conversation!
Last week I read a Substack post by the editor of a literary magazine. The post was about submissions strategy and how to interpret certain forms of rejection from journal editors.
I wanted to link to the piece in Monday’s news round-up. But I could not, in good conscience, do so. I was certain the piece was heavily edited with the use of AI.
The many issues surrounding AI and its uses are complicated. Some writers may not use it at all, some may use it strictly for research, others might use it for editing and proofreading. I’m not interested in policing other people’s uses of the technology.
I do, however, draw a red line at sharing articles so clearly written with the help of AI assistants that they possess what I think of as “AI voice.” What is AI voice, exactly? I’ll get to that in a moment.
Suffice it to say, I draw an especially bold red line at articles in AI-voice, produced by editors of literary magazines. The role of lit mag editors is ostensibly to curate, preserve and protect quality literature. What does that mean? It means, among other things, seeking out and showcasing fresh voices. New voices, different voices, interesting voices. Indeed, among the hundreds of editor interviews I’ve conducted, a vast majority have said that the single thing which moves a submission from consideration to acceptance is a strong, unique voice.
Given the proliferation of AI-generated writing, one might contend that the role of lit mag editors to identify and showcase fresh human voices is more important than ever. Thus when I encountered this essay by a journal editor, clearly crafted in AI-voice, not only was I taken aback but I drew a firm line against sharing such pieces here.
I also reached out to the editor to determine if my suspicion of AI-use was correct. And here, as an AI-bot might say, is where it really gets interesting. Upon reaching out, what followed was one of the most touching and surprising conversations on the subject of AI-assisted writing I’ve ever had. I will tell you all about it.
First, I want to take a closer look at AI-voice. It’s come to my attention that many people do not know its attributes. I believe we in the literary world should all take a moment to learn how to recognize it.
For this we will need some help from actress Drew Barrymore.
I recently came upon a Facebook post from a group called This Day in History. They highlight stories about celebrities and other famous people. I am fairly confident that all their content is AI-generated. The fact that none of their pieces have authors is a key sign of this.
One of their recent posts is about Drew Barrymore. Give it a read. Don’t worry—it looks longer than it feels. The piece is juicy, interesting, inspiring, and you’ll learn a thing or two about the child star.
Can you spot the AI tells?
At seven years old, she made the world fall in love with her.
E.T. the Extra-Terrestrial turned Drew Barrymore into a global star before she was old enough to understand what global meant. The gap-toothed smile. The unguarded wonder. The complete, natural openness that made audiences feel they were watching a real child encounter something magical — because they were.
What they didn’t see was what happened after the cameras stopped.
Drew was born into the Barrymore acting dynasty — a family name that carried the weight of generations of fame and the specific kind of dysfunction that tends to travel alongside it. Her father struggled with alcoholism and largely abandoned the family. Her mother, a struggling actress who saw her daughter’s stardom as a second chance at relevance, made choices that no child should have been subject to.
When E.T. made Drew a global star at seven, childhood disappeared overnight.
At nine, she was taken to Studio 54 — the epicenter of celebrity excess at its most unapologetic. At nine, she began drinking. At ten, marijuana. By twelve, cocaine.
“I didn’t have parents,” she later said. “I had enablers with checkbooks.”
She was photographed on red carpets while privately unraveling. She smiled for cameras while the world that was supposed to protect her did the opposite.
At thirteen, the spiral became impossible to manage. She was placed in a psychiatric facility — not a celebrity retreat, not a discreet private clinic, but a locked ward. She spent eighteen months there. She has described it since with the unflinching honesty that would become her signature: it was harsh. It was also the most important thing that happened to her. It was the first place where someone treated her survival as worth working for.
When she was released at fourteen, she made a decision that most adults would find terrifying.
She legally emancipated herself from her mother.
At fourteen years old, Drew Barrymore became legally responsible for her own life. No guardian. No safety net. A teenager, alone in Los Angeles, with a famous name and almost nothing else.
Hollywood, predictably, wasn’t sure what to do with her. Child stardom is fragile. Public addiction makes it catastrophic. She was seen as a liability. Studios hesitated. Insurance companies hesitated. Roles dried up.
She kept going anyway.
Small parts. Auditions. Rejections. The slow, unglamorous work of rebuilding a career that had been built on a foundation that couldn’t hold her weight. She didn’t disappear — she just kept showing up, in smaller rooms, with less certainty, waiting.
Then came 1998 and The Wedding Singer. Audiences saw something in her that they recognized — warmth, resilience, the particular quality of someone who has been through real things and come out the other side with their humor intact.
They fell in love again. This time not with a child. With a woman.
In 1995, at just twenty, she co-founded Flower Films. She didn’t want to be a product of Hollywood anymore. She wanted to shape it. Charlie’s Angels. Never Been Kissed. 50 First Dates.She wasn’t just starring in films — she was producing them, influencing every element of their creation, moving from someone the industry controlled to someone who controlled her own space within it.
She wrote her first autobiography at fifteen. She has never stopped being honest about where she came from. The addiction. The institutionalization. The emancipation. The long, imperfect, ongoing work of building a stable life from genuinely unstable materials.
Today, she hosts a television show marked by the warmth and emotional transparency that have become her defining qualities. She is a mother to two daughters and has spoken often about consciously choosing to be different — to break the cycle, to be present, to be the thing she needed and didn’t have.
The real achievement — quieter than any box office number, more durable than any award — is that she learned how to raise herself.
She was nine in nightclubs. She was thirteen in a locked ward. She was fourteen living alone in a city that had already used her up and was ready to move on.
She did not just survive that. She studied it. She understood it. She used every scar as material for becoming something the chaos of her early life never intended her to be: whole.
Yay, Drew! You go, girl!
So, what were AI tells?
Hyperbole: “At seven years old, she made the world fall in love with her.”
LLM’s frequently use extreme language and hyperbole. As the Wikipedia entry for Signs of AI Writing says, “LLM writing often puffs up the importance of the subject matter…”
Narratives collapse. Meaning implodes. Normalcy craters. Childhood disappears overnight. And a seven-year-old made the world fall in love with her.
Could a human have written this? Yes, sure. But right out of the gate, your AI spidey senses should spike here. It is simply false to say the entire world ever felt one way about anything, let alone Drew Barrymore. But the line grabs attention and builds a story of epic magnitude.
Sentence fragments, the rule of three, and em-dashes. “The gap-toothed smile. The unguarded wonder. The complete natural openness that made audiences feel they were watching…something magical—because they were.”
This section is a clusterfudge of AI tells. Here are punchy little sentence fragments, a list of three, and an em-dash, all at once!
From Wikipedia:
LLMs overuse the ‘rule of three’. This can take different forms, from “adjective, adjective, adjective” to “short phrase, short phrase, and short phrase…LLMs often use this structure to make superficial analyses appear more comprehensive.
Yes, the rule of three, sentence fragments and em-dashes are all writing techniques of human beings. And yes, it is entirely unfair that these wonderful linguistic devices have been pilfered by AI bots. Nonetheless, this is our present-day situation.
“Not this, that” sentence construction: “She was placed in a psychiatric facility—not a celebrity retreat, not a discreet private clinic, but a locked ward.”
This is one of the biggest AI tells. Once you notice it, you will see it everywhere. The construction is called a negative parallelism.
As a rhetorical device, negative parallelisms are quite effective. The narrator wishes to emphasize the uniqueness of the situation. This wasn’t one thing you’ve heard about; it was something far different. By setting up what it could have been, the narrator emphasizes the importance of what it was.
It’s effective and compelling, and it is so very AI.
Setting up the reader for a dramatic twist: “What they didn’t see was what happened after the cameras stopped.”
This sentence contains two AI tells. The first is the way the bot has set up a high-stakes situation. The entire world is in love with this girl. But they missed something—the real girl behind the cameras. The girl the narrator will share with you now. It’s similar to another rhetorical device frequently seen in AI-writing: “Here’s where things get interesting.” Or, “Honestly?…”
The second tell here is the phrase “after the cameras stopped.” My ear picks up on this as being choppy and awkward. AI bots like to abruptify and punchify sentences. The cameras didn’t stop rolling, taping, filming. The cameras just stopped.
Using opposites to emphasize a point: “She didn’t disappear—she just kept showing up.”
This is a form of the aforementioned negative parallelism.
A human might say simply, “But Barrymore persisted.” Or “But the young actress would not be deterred.”
Instead, AI bots tend to build intrigue through opposing ideas. She didn’t do X, as you’re probably thinking. She did Y, as I am revealing to you now.
A real truth: “The real achievement—quieter than any box office number, more durable than any award—is that she learned how to raise herself.”
AI prose is often inviting you to lean in, to hear the real truth that it’s about to deliver.
Redemption arc. “She wasn’t just starring in films—she was producing them.”
The piece ends with Barrymore reclaiming herself not as a child star, but a full-grown actress. She embraced comedic roles and then began to produce and direct.
AI pieces often contain a satisfying and hopeful shape. Obviously, this is also true for human-written pieces. But in pieces that are written by or heavily edited by AI, there is almost always a shift toward healing and redemption. The story is compact, tidy and neat.
Ideas that don’t make sense. “She used every scar as material for becoming something the chaos of her early life never intended her to be: whole.”
This one is sometimes trickier to see because on quick reading, you grasp the idea. Barrymore’s scars made her stronger, etc. But upon analysis, these words don’t quite make sense. How could it be that “the chaos of her early life never intended her to be” something? Does chaos have intentions? Does chaos care whether Drew Barrymore is whole or not? Does chaos intend for things to happen?
Read it quickly and you grasp it. But stop to think about it and, well, the narrative collapses, meaning implodes, normalcy craters—it’s gobbledygook.
Did you find other AI tells? Please share them.
Now, let’s get back to our editor.
When I reached out, I wasn’t sure what I was expecting. I had a feeling I’d get no reply at all.
To my surprise, she wrote back right away. That she published such a piece had been on her mind too. In fact, I was not the first person to inquire about the article. She admitted to me that she “got a little over-zealous with the AI-powered ‘clarity’ suggestions in the editing software I’ve used since before AI-times (Grammarly).” She also acknowledged that “It polished the life right out of my prose…”
The editor went on to tell me that she’d been “thinking a lot about my instinct to polish my awkward and weird voice and the ‘let-me-give-you-what-you-want-engine’ that AI offers.” But the message she received from another reader had reminded her that what people actually want to hear is her own voice—“my weird awkward natural voice.”
In an additional note of graciousness and humility, she added, “I appreciate your note as a second reminder that I’ve gotta be me. (Sad that this is often my advice to writers and I didn’t follow it.)”
I was deeply touched by this editor’s transparency and thoughtfulness. We went on to chat more about human voice, AI voice, and the voices in our heads that fill us with doubt as we try to work. To my added surprise, the editor decided to pull the piece and then re-post it in her own voice. It’s live now and it’s a million times better—softer, realer, more human. I will share it in a news roundup sometime soon.
My conversation with this editor got me thinking about a feature of AI writing assistants I’d not before considered. We’ve often discussed how they encourage laziness, how they disable critical thinking. We’ve considered the plagiarism angle and other ethical concerns.
One thing I’ve not seen much talk of is how AI-assistants dampen one’s own voice.
Finding your voice as a writer is hard work. Where does it come from? What shapes it? How do you even find it to begin with? How do you know if you have a unique voice? Does every writer have a unique voice? Do writers need a unique voice in order to tell their stories and be successful? What is voice?
These are not simple questions to answer.
But I do know this: Relying on AI-assistants to help craft one’s prose is one way to ensure that a writer never discovers their own voice.
It is tempting, very tempting to use these tools. One moment you’re asking AI for police procedure for a car theft in Kentucky. Then you’re asking AI to describe the kinds of accents people in Kentucky have. Then you’re asking it to proofread a scene from your Kentucky crime story. Then you’re accepting AI’s suggestions to make the scene sound truer to Kentucky life. Then you’re letting the bot write full sentences for you and inserting them into your own prose—hey, it actually sounds pretty good, why not, it’s just a sentence here and there.
But I believe there is a cost. And that cost is voice. Voice. One of the most potent tools we have as writers. One of the most potent forces we have, as people.
AI-assisted writing turns every essay into an essay with the exact same prose style. The writer’s voice gets lost, subsumed by uniformity and conformity. Worse yet, the very process of relying on AI steers one away from finding one’s own voice, from digging for the unique word combinations that are true to oneself, from finding a singular language to express what is deeply personal or deeply urgent, deeply true.
AI can help us all churn out prose faster. It can pump out content for us with just a few clicks.
But it cannot help us find our own voice. In fact, I believe it actively hinders us from doing so.
And what a loss that is, for literary culture, for writers, for humanity. If we do not have our voices, after all, how do we ever intend to use them?
I’m so thankful to this editor for her honesty, transparency and willingness to talk. I think it’s up to all of us to have such conversations and to determine where our red lines are when it comes to AI-writing tools.
It’s up to all of us to protect and nurture our own real, human voices.
If writers and lit mag editors do not take up the cause—honestly? Who will?
What do you think?
Are you aware of the common tells of AI? Does AI-voice get on your nerves?
Will this guide help you recognize AI-assisted prose?
Does reliance on AI for writing and editing smother a writer’s voice? Should we be concerned about that?
Don’t you just love Drew Barrymore?
Voice is one of the things I look for in submissions, and one of the things I find to be rare. Perhaps I receive my share of AI generated work, but even among what I believe in good faith are human writers, my most frequent reason for declining work is that the voice is indistinguishable from others. That is to say, I recognize that the piece is done well, but the voice of it doesn't stand out.
I'm fully in favor of writers using their own human voice, and avoiding AI altogether. However, there's an alarming new trend emerging. I don't know how widespread it is yet, or if it will indeed take hold, but in the fear that their work might be considered AI, some authors are deliberately making their work less articulate, and deliberately leaving in errors: in vocabulary, grammar, and punctuation. They're simplifying (and smoothing down) their work in various ways so they appear less intelligent, less capable of writing coherently. In a great irony, some use "dumb-down" programs that help identify and remove (or change) parts of the text that might be considered AI built.
Anybody doing this has already lost their way. They're writing in anticipation of what people might think. We already have enough of that in the form of subject matter that's avoided, certain words that people are afraid to use, and the fear of simply calling a spade a spade. Basically, it boils down to the fear of offending someone, somewhere.
Now some want to add in the fear of sounding like AI? Please folks. Don't go there. Don't hesitate to use an em-dash if you want. Don't worry about saying "when the cameras stopped" (we already know what cameras do, no need to spell it out). Don't let all these so-called "tells" creep into your rule book. If we do that, instead of nurturing our own individual voice, we risk falling into a whole other trap, the one in which we're trying too hard to make sure everybody knows AI wasn't involved.
Hey, go Drew! These kinds of mini-biopics are all over the Web, hitting the dormant (or not so dormant) fan in us. Yes, AI will rob you of your style, your quirks, wash it with chlorine. Did I just do a rule of 3? What really bugs me is that we are being "polluted" even if we don't use AI at all. Reading that engineered slop leaves little traces in our minds. I don't want to start writing like a bot but they are everywhere. AI is an invasive species. I don't want my mind to goooooo. Going to go read a little Henry Miller to set me straight.