How Kindle novelists are using OpenAI’s ChatGPT

Earlier this year, I wrote about genre-fiction authors using AI in their novels. Most wrote for Amazon’s Kindle platform, where an extremely rapid pace of publishing, as fast as a book a month, is the norm. AI helped them write quickly, but it also raised complex aesthetic and ethical questions. Would the widespread use of AI warp fiction toward the most common conventions and tropes? What parts of the writing process can be automated before the writing no longer feels like their own? Should authors have to disclose their use of AI?

With the debut of ChatGPT, many of the questions these writers were dealing with have become more urgent and mainstream. I checked back with one of the authors, Jennifer Lepp, who writes in the cozy paranormal mystery subgenre under the pen name Leanne Leeds, to see how she was thinking about AI now. She’s still using the GPT-3-based tool Sudowrite — in fact, she is now paid to write tips on using it for the company’s blog — and has begun incorporating some of the more recent tools into her fiction. We spoke about what it’s been like working with ChatGPT, how its debut has roiled the independent author community, and other topics.   

When we spoke last time, you had gone through an evolution of using Sudowrite, first mostly as a sort of thesaurus, then experimenting with incorporating its text into your work, then letting it lead you and having an alienating experience with that and reining it back and using it primarily to flesh out descriptions you’d outlined. What’s your process like now?

Well, I had hoped that it would help me write two books at the same time, and that failed spectacularly. Apparently, I’m still connected to my own writing. So, on the one hand, that was good.

You thought that it could allow you to toggle back and forth and write two books simultaneously? 

I figured, Hey, if I don’t know what to write, I’ll just pop something in there and it will get me going, and I’ll be right back into the book I left a week ago. It didn’t quite work out that way. If I didn’t know what I was doing, it didn’t matter what it spit out at me. It wasn’t going to help me reconnect with material I already wrote.

You and a few other independent authors were early adopters of these tools. With ChatGPT, it feels like a lot of other people are suddenly grappling with the same questions you were confronting. What’s that been like? 

I definitely am still grappling, and I think I’m grappling a little bit more publicly. For the most part, people before had kind of rolled their eyes — I don’t think they understood what people were using AI for. ChatGP3 exploded that. Every group, every private, behind-the-scenes author group I’m in, there’s some kind of discussion going on.

Right now, everybody’s talking about using it on the peripherals. But there seems to be this moral chasm between: “It does blurbs really well, and I hate doing blurbs, and I have to pay somebody to do blurbs, and blurbs isn’t writing, so I’m going to use it for blurbs.” Or “Well, I’m going to have it help me tighten up my plot because I hate plotting, but it plots really well, so I’m going to use it for that.” Or “Did you know that if you tell it to proofread, it’ll make sure that it’s grammatically correct?’ 

Everybody gets closer and closer to using it to write their stuff, and then they stop, and everybody seems to feel like they have to announce when they’re talking about this: “But I do not ever use it its words to write my books.”

And I do. It doesn’t drive my plot. It doesn’t generally drive any of the ideas in my books. It doesn’t create characters. But the actual words, just to get them down faster and get it out, I do. So I’ve found myself in the past couple of weeks wondering, do I engage in this debate? Do I say anything? For the most part, I’ve said nothing.

What do you think the line is that people are drawing?

It’s a concern of plagiarism. Everybody knows that they crawled stuff with permission and without permission. 

And there’s an ethical question. I can go in and — right now, I’m listening to Jim Butcher’s audiobooks. I love his tone. I love the deadpan snark. So I went into the AI when I was thinking about trying to get something like that with a character and said, “Rewrite it in the style of Jim Butcher.” Bam! The same kind of deadpan, urban fantasy phrasing. 

Well, where did it get that? It’s almost exactly the same argument and the same fear that’s going on with visual artists. It’s just much more obvious in the artist community. I have three authors that I’ve read extensively, indie authors that I’m friends with, and I know they never gave permission for their stuff to be looked at, and I was able to reasonably recreate their style.

Do you see a line between using AI for something like a description and using it to mimic another author’s voice?

Yeah. That I won’t do. That, for me, is an ethical line. I may like Jim Butcher, and I may wish to God I could write like him, but I’m not going to take my stories and have them rewritten in his voice to rip him off. 

But you could, if you were ethically okay with that, with this technology and what it allows you to do.

Have you incorporated ChatGPT into your work? 

Right now, I use it for titles and plots — specifically mystery plots. And blurbs.

I basically started out by just telling it who I am and what I need. “I am writing a paranormal mystery that takes place in the small town of Table Rock, Texas. It has a female amateur sleuth. This is her name. I need a murder victim. I need how they were killed. I need four murder suspects with information about why they’re suspected and how they are cleared. And then tell me who the guilty killer is.” 

And it will do just that. It will spit that out.

What are some of the things that it’s given you?

Right now, I have [plots for] books two, three, four, five, six, and seven, and all of those murder mysteries were ChatGPT-generated, though I edited some of them. The impressive thing about it is that if I tell it that it’s a cozy mystery and I tell it that it needs to be humorous, it seems to understand what I’m asking for. The names that it gives me for the suspects are cutesy. The reasons behind it are never gory or serious.

You feel like you can automate that part of it and still feel in control of the story? 

There are two parts of a cozy mystery. There’s the murder, and the murder is the thing that all of the characters revolve around. But the murder, to me, tends to be less important than all the revolving. So there has to be a murder, and it has to be amusing and funny and give reasons for mayhem and strangeness. But what it is is almost inconsequential to the plot, even though it’s the thing that drives everything.

You mentioned over email that you were using AI for book covers.

I didn’t do the whole cover on DALL-E, but on the seventh book that I had, I had sketched out a plot that involved a Lykoi cat. It’s a cat that is so ugly it’s cute. It’s apparently a fairly new breed that was like a crossbreed between a cat with hair and a hairless cat. And so it’s got hair in some places, and it looks like a werewolf.

So I would have had to find a photographer that could do a shoot, find a Lykoi cat, pay everybody to get the image and the cover that I needed. That’s expensive. So on a lark, I was like, Huh? I wonder… 

And I went to open my account, jumped into DALL-E. Boom! For me, it saved so much time and money, and the cover looks great, but a photographer didn’t get paid, right? Somebody who wanted to pose their cat didn’t get paid.

How do you see these tools and the way writers use them evolving? 

I’m really just stuck in the middle, wondering which way it’s going to go. I definitely don’t want to encourage people who aren’t comfortable using it to use it. I do think it’s going to leak into their lives. It’s already leaking into all our other software, so I think it’s going to be very hard to get away from. But I definitely don’t know where it’s all going. ChatGPT shocked the hell out of me. I had thought, well, it’ll take three or four years, and it’ll get better. Then came ChatGPT, and oh my god, that’s so much better! It’s been six months! The progress is so incredibly fast, and so few questions have really been answered.

The interview has been condensed and edited.