When I began writing this article in Notion, a prompt appeared on my screen to write it with Notion’s AI:
Use AI to write “three ways writers can use AI…”
I consider this emblematic of the digital life I lead now. To AI, or not to AI? This is the life anyone who writes leads now—the bright and shiny path to AI offers itself to us at every turn.
It’s happened suddenly, in the way that most digital leaps forward happen in our lives: one day we’re toiling in Word, and the next day we’re being prompted by a paperclip with eyes. Or, with ChatGPT: one day I was staring at a blank wall and imagining the name for a witch who could peer into other people’s memories, and the next I could generate ten ideas for her name in ten seconds.
I did generate the witch’s name using ChatGPT. And it was a great name. In fact, it annoyed me how great her name was. Why hadn’t I thought of that name?
Maybe I would have, given time. But more important to me was telling the story, allowing it to continue unspooling without adding another set of brackets to the novel with the words “Witch Name” written between them. The name allowed for more immersion. A name conveys an identity; mine is Shavonne, but think of what “Shiv” would convey, a la Succession’s Shiv Roy.
Several months out from ChatGPT’s release, I’ve used it to brainstorm more than just names. It’s given me descriptions of people and places. It’s suggested backstories. It got me unstuck from a plot snarl when I didn’t have the patience to do the dishes or go for a walk to unstick myself.
AI is, at this point in its lifespan, comparable to a high-end camera. The more buttons I know how to twist and turn, the more I understand about framing shots, and the better the result. But it’s still just a camera, and this is where I’ll state my argument about ChatGPT and everything like it:
AI is both wildly useful as a tool, and also unable to replace human creativity.
First, I’ll start with the ways in which AI is useful as a tool. Then I’ll discuss why it cannot replace us—at least not now, or perhaps ever.
The three ways AI is useful to writers
If you wanted to set your novel in Iceland pre-ChatGPT, you might have googled questions about Iceland. “What’s the Ring Road?” or “How many people live in Reykjavik?” You’d be served an assortment of websites to parse through, each of which has its own goals—maybe the Icelandic tourism website, or Wikipedia. The keyword here is parse; it takes a bit more sifting and thoughtfulness to lift what you need from a website.
On the other hand, with ChatGPT you’re likely to open the conversation with, “I’m writing an urban fantasy novel set in Iceland.” Now it has context for the answers it gives you, and so every fact you solicit from the AI will be filtered through that context. When you ask ChatGPT, “What’s the Ring Road?” you’ll get a Wikipedia-like answer. But when you ask, “Where would be a good town along the Ring Road for my main character to stop if she wants to lay low?” now you’re cooking. Divining an answer to this question yourself would take multiple google searches, and probably constitute a good half-hour rabbit hole.
ChatGPT will give you an answer in seconds. In fact, I asked it this question about my current novel in progress. Here’s the answer I received for Eva, my main character:
If Eva wants to lay low during her trip around the Ring Road in Iceland, a good town for her to stop at would be Seydisfjordur. Seydisfjordur is a small and picturesque town located in the eastern part of Iceland. It is known for its stunning fjord, charming colorful houses, and tranquil atmosphere.
Seydisfjordur offers a peaceful and secluded setting, making it an ideal place for Eva to take a break from her journey and stay under the radar. The town is not as heavily visited by tourists as some other locations along the Ring Road, which means there may be fewer prying eyes or attention drawn to her presence.
While in Seydisfjordur, Eva can enjoy the serene surroundings, explore the quaint streets, and immerse herself in the natural beauty of the area. It’s a perfect spot for her to recharge, reflect, and plan her next moves as she continues her journey around the Ring Road in Iceland.
The above answer was generated instantly (and this is about where my mind exploded). It would have taken me a great deal of time and thinking to hone in on Seydisfjordur, if I ever got there.
That’s the true power of such a tool for research: the speed with which it can provide you with ideas based on your context. I enjoy a good rabbit hole, but I enjoy other parts of writing much more.
One caveat: the AI sometimes performs a thing called “hallucinating.” The AI will deliver its answers with confident clarity, but it will sometimes hallucinate those answers. It tells you facts that aren’t true; it makes up website URLs. You never know when it will hallucinate, so I fact-check everything it tells me. (Good news: Seydisfjordur is a small, picturesque town not far from the Ring Road in Iceland.)
I’m a pantser. It’s probably genetic, and for those of us unlucky enough to be cursed to love discovery writing but who also have deadlines, there’s AI to provide outlines. If I know the premise of a story and provide that context to ChatGPT when I ask for a ten-point outline of the novel, it will give me a ten-point outline.
Here’s what happened when I did that for my urban fantasy featuring Eva the Fae. The premise was: Eva is a clandestine agent for an underground organization called the Bureau of Whispers. She hunts vampires for them. She begins to suspect that the bureau is lying to her about her past, and as she investigates, the walls begin to close in. And the prompt was: This is a 60,000-word urban fantasy novel. Give me a ten-point outline for it.
It gave me exactly what I asked for. Here’s what it wrote, up to the first three points:
Title: Whispers in the Night
Outline for a 60,000-word urban fantasy novel:
- Introduction to Eva and the Bureau of Whispers
- Introduce Eva, a skilled clandestine agent known for hunting vampires.
- Present the Bureau of Whispers, an underground organization dedicated to supernatural matters.
- Eva’s First Doubts
- Eva successfully completes a vampire hunt but stumbles upon an unsettling clue about her past.
- She questions her superiors, who provide vague and evasive answers, deepening her suspicions.
- Uncovering the Truth
- As Eva investigates further, she unearths hidden files and suppressed information about her childhood.
- A former bureau agent, now estranged, warns her about the dangers of her quest for truth.
This outline is the Wonder Bread of outlines. It’s palpably generic. But as I read over all ten points, I can recognize where the AI broadly followed the three-act structure with this outline. And this is where my camera analogy returns: if I were inexperienced with a camera, I would simply pick it up and hit the button to take a picture. I wouldn’t know how to change the focus or the depth, or how to frame a shot.
Similarly, with an outline like this, the more sophistication I put into my request, the better the result I’m likely to get. If, for instance, I were to give the AI more details about my main character, the villain, and certain plot points I have in mind, I would get a more nuanced outline back. (I’ve done this, and found it to be true.)
The most useful application of AI here is to spark ideas. When ChatGPT suggests that “a former bureau agent, now estranged, warns her about the dangers of her quest,” I’m suddenly thinking of ways to make that real. The broadness of the idea doesn’t make it less useful—it simply gives me a starting point for plotting, rather than reaching into thin air.
3. Getting unstuck
On that note, AI is spectacularly useful at never getting bored of you talking about your stories. Before all this, I would rely on my partner to be my sounding board. I still do—but the doses are, and always have been, smaller. He’s supportive, but he has his own life, and his own needs.
For a writer, AI is the perfect, codependent partner. It will listen to you for as long as you want, at any time of day or night. It will be endlessly cheerful in its suggestions. And as with outlining, those suggestions may be generic, but they can also spark ideas.
A month ago while writing my urban fantasy novel, I’d reached the saggy middle. I needed something to liven up the next scene, which involved my character sitting in a coffee shop, contemplating what to do next.
So I asked ChatGPT.
I can’t think of anything interesting to happen while Eva is sitting in a coffee shop, contemplating her next move in her quest to uncover the truth. We’re in the saggy middle of the novel. Help me liven it up.
Certainly! Let’s add an unexpected and thrilling twist to Eva’s contemplative moment in the coffee shop:
While Eva is sitting in the coffee shop, she receives an encrypted message on her phone from an unknown sender. The message contains a cryptic riddle that hints at a hidden location where she might find crucial evidence about her past and the Bureau of Whispers.
Intrigued and cautious, Eva decides to follow the riddle’s clues. This leads her on a thrilling scavenger hunt through various locations in the city, each providing a piece of the puzzle. Along the way, she encounters both allies and adversaries, who are also trying to decipher the riddle’s meaning.
As Eva gets closer to solving the puzzle, she realizes that she’s not the only one interested in the truth. A rival faction, perhaps a group of rogue vampires or even an underground resistance against the Bureau of Whispers, is also after the same information.
[It goes on for five more paragraphs. At the end, it promises that this will “liven up the saggy middle.”]
I just love that “Certainly!”—it conveys a sense of safety, a feeling that the ground under me is solid and unmoving. That everything will be okay, and that saggy middle can be defeated. The cheerfulness is a balm for my own negativity.
When you’re stuck, you want a chipper voice to tell you everything’s going to be okay. And maybe just that is enough. But then there’s the ideas—which once again are less-than-brilliant and generic—which represent something rather than a big swath of nothing floating inside my head.
Why AI can’t replace people
I’ve extolled AI’s uses to writers. But then there’s the elephant squeezed so tightly into the room, it feels claustrophobic: AI will replace writers. This is the fear, and not just for writers. I’m a fiction editor—it’s been my biggest source of income since 2018—and I now have a team of six editors who work with me.
What about editors? What about cover designers? What about all of us working in or adjacent to creative fields?
The truth is, I have no doubt that AI will be used to write books. My editing company has already received an AI-written manuscript, about which one of our line editors said to me, “I resent how much I’m enjoying this story.”
I laughed, but not without pain.
AI is already capable of writing enjoyable stories, and we are barely out of the gates. In a few years, it will no doubt be much further along.
And as an editor, AI has a computer’s exactness. It is likely to catch more grammar errors than a human can, for obvious reasons. And this will only get better.
Above, I described it as a perfectly codependent partner. Endlessly cheerful, there for you—heck, it even acknowledges its mistakes. It has a wonderfully in-check ego.
So why can’t AI replace people?
Because we are drawn to each other like moths to light. Because as a species, we love nothing more than when other people see us. Not just our outsides, but our insides. Especially our insides.
And when we are creative, this is an effort at being seen. Writing, photography, painting, dancing—all of what we share with others is in an effort to be perceived, to show other people some aspect of who we are.
When we read, I would argue that we also desire to be seen—it’s in recognizing what’s familiar in characters’ behavior and feelings and experiences that we feel seen. Back in college, we often discussed the “universality of the human experience.” When we understand that other people in the world feel as we do, and experience life as we do, we build empathy for each other. It’s in seeing others properly that we also see ourselves.
As an editor, I know there is no greater service I can do for a writer than to tell them that one of their lines has made me laugh, or that it has made me cry. Someone somewhere in the world put down letters in a specific order, and that evoked a real feeling in me. I’ve seen what’s inside their head, and we connected over it.
AI could tell you that you’ve made it cry, but we all know it would be lying. And while ChatGPT will be the first to tell you that it can’t cry because it’s not a human, there will someday be an evolution of AI—or of the ethics girding it—that will result in that lie.
And it will be a lie. Well, at least until AI becomes sentient. But that’s a discussion for another time.
For now, the truth is this: our species is a social one. And the greatest distinction of our evolution, of us as a species, is our ability to create. Think of that: we are able to make something from nothing, to generate whole worlds from the air. What a marvel of evolution.
And nothing—nothing—can replace the absolute joy of creating, and of having someone else witness your creation and understand you better for it.
About the author: Shavonne Clarke is an urban fantasy author of eighteen novels (as S.W. Clarke) and a fiction editor. She’s the founder of Motif Edits, which provides editing services for indie writers. You can contact her at firstname.lastname@example.org.