In short
The publishing industry’s AI scandal and whether readers can tell AI-written books from human writing, plus what “guardrails” could prevent harm.
Guests and backgrounds
Imogen Westnight, journalist-novelist for Slate; Wahine Vara, writer/journalist (Businessweek, The Atlantic, The New Yorker), author of Searches.
Key claims
Readers and outlets alleged Mia Ballard’s self-published horror Shy Girl was AI-assisted; Hachette pulled the book after NYT investigation. AI detection software is unreliable and may wrongly accuse authors. AI text isn’t inherently “soulless,” and people often can’t distinguish AI from human prose when style is trained on an author.
Notable examples
Shy Girl’s “flat” prose and repeated adjective/simile patterns; NYT used AI-detective software; a YouTube critique by Frankie’s Shelf; Vara’s experiment where close friends couldn’t identify AI vs her own passages, including an AI passage with a grammatical error that humans missed.
Written by AI. May contain mistakes. Listen to the episode to check what was said.
Chapters
Tap a time to open that second in VOThe Story of 'Shy Girl'
2:14 to 5:48
Overview of the self-published horror novel and its rise in popularity.
“During the course of this incarceration, she starts to take on kind of animal qualities and finds herself actually seemingly turning into a dog.”
AI Accusations and Reactions
5:48 to 9:38
Discussion on the allegations of AI involvement in the writing of 'Shy Girl'.
“has issued very sort of sparse statements about this.”
Impact on Author and Industry
9:38 to 10:46
Exploration of the consequences faced by the author and the publishing industry.
“But then it ought to have been picked up when it went through Hachette.”
Exploring AI's Impact on Writing
17:16 to 19:32
Wahine discusses her experiment distinguishing her writing from AI-generated text.
“So I think there's this, I don't know, misconception among writers, among readers, that there is a certain kind of way that AI generates language, and it's super different from the way writers do.”
AI's Grammatical Precision vs. Human Flair
19:32 to 22:26
Wahine and the host compare AI and her writing, revealing AI's strengths and weaknesses.
“So the people who know you best in the world don't know you that well, apparently, or AI is exceptionally good at what it is doing.”
Public Perception of AI in Literature
22:26 to 26:53
Discussion on how AI-generated texts are perceived by readers and implications for literature.
“This is the thing that I would like us to talk about.”
Transcript
Automatic transcript. May contain errors.0:00Noel King:AI use and allegations of AI use are causing chaos in the publishing industry. Some authors have had their books picked apart and their careers destroyed. AI scandal that is shocking the book industry. If you're wondering if it's AI slop, it is AI slop. AI accusations are no fucking joke. Some writers say AI is just a tool and they feel no shame about using it. I'm Coral Hart and I write AI-powered romance novels that are topping Amazon charts. So is it that AI is soulless and could never replicate what humans do? Or is it pretty compelling? Coming up on Today Explained, a scandal, an experiment, and a question.
0:39Noel King:Can books survive the AI revolution?
0:59Noel King:Support for Today Explained comes from ServiceNow. AI is moving fast, but without visibility, it's just chaos. Different tools, different models, different teams using AI in completely different ways. A mess. ServiceNow turns that mess into control. With the AI control tower, you see all your AI across the business in one place. Oh, what it's doing, what it's done, what it's about to do, so you stay in control. To put AI to work for people, visit servicenow.com.
1:54It's time for a trip to Ross. Work your magic.
2:01Butterfly in the sky. I can go twice as high. Take a look. It's in a book.
2:09Noel King:AI.
2:13Noel King:Imogen Westnight, journalist-novelist, wrote about a recent massive blow-up in the publishing industry for Slate. so what happened was an author called Mia Ballard an American author self-published a horror novel called Shy Girl and the conceit of this book is it is about a woman who is down on her luck financially and decides to sign up for a sugar baby website in the hopes of fixing that she gets put in touch with this man and enters his orbit and eventually he um sort of kidnaps her and decides to keep her as a pet like a dog in a sort of cage and on a collar and forbids her from speaking except to say woof and not to woof to actually say woof which is a bit strange but there we go this is who i am now a pet a shape carved by someone else's hands a thing devoured piece by peace until there is nothing left but obedience, the quiet and the hurt, until hurt is all that remains.
3:15During the course of this incarceration, she starts to take on kind of animal qualities and finds herself actually seemingly turning into a dog. I chose the man who wanted not who I was, but who I could become, a pet, a prisoner. She self-publishes this book and it's popular among fans of self-published horror online. And then what happened was, as often enough happens these days, is that a big traditional publisher, in this case Hachette, will see a self-published novel and see that it's popular and decide to then publish it again under the imprint of their own house. And that can be a pretty good deal for the house and for the author because they get a book and a payment and the publishing house know that people like this book already and therefore it's likely to earn a profit.
4:07So they do this, they contact her. It then gets published in the UK by Hachette and is slated for publication in the US a little bit later than that. People begin to read this book more widely now that it's out with a major publishing house and And discussion begins to pop up on places like BookTok. The writing felt really flat. It felt surface level and underdeveloped, which could be a clear sign of AI usage. I was just reading sentences and I was like, that doesn't even mean anything. And read it. Shy Girl by Mia Ballard. Does anyone else think this was written by ChatGPT? It has this very recognizable and constant rhythmic use of adjectives and similes that stinks of AI.
4:56this conversation kind of gathers speed and people agreeing that there's something kind of off about the book and then it all gained a little bit more attention there was a user a youtube book person called frankie's shelf who made a three hour long video dissecting this book and all the things that they thought um gave the signs of being ai generated it's so empty. It's flat in every way. Themes, characters, plot, writing. And that is because, in my opinion, large chunks of this book were written with assistance from a generative AI. The New York Times then get involved and decide to run their own little investigation into the book, doing things like putting it through AI detective software, and take that to Hachette.
5:43Hachette then get off their ass, as it were, and pull the book. And then the author Mia Ballard has issued very sort of sparse statements about this. She gave a quote to the New York Times saying, you know, that this is essentially ruined her life and her reputation. But the interesting claim that she made was not that there was no AI in the book, but that potentially there was AI in the book because she had hired or asked a third party to edit the book.
6:12Noel King:What do we know about the author? This is her second book. She published another, just self-published. And we don't know loads and loads about her because she's very much withdrawn from the public eye in the aftermath of all of this. An extra layer to the narrative around all of this is that she's a black woman and therefore someone who is traditionally less interesting to big publishing houses and is afforded less time and attention. It would be silly to argue that none of the vitriol is linked to the fact that she's a black woman. You know, I think there's like one ought to be careful about the way that we talk about this because this is someone from like a from publishing terms anyway, certainly like a disadvantaged background who's been given this big platform and has attracted like a lot of flack for this potential use of AI.
7:09Noel King:You read the book. For your money, what made it so obvious that AI had been used? There's like two layers to it. Everybody is becoming familiar with certain kinds of AI writing tells, right? There's things like negative parallelisms, that thing where it's like, it's not just this, it's this. Or excessive use of metaphor and similes, especially ones that don't quite make sense or that come very rapidly one after another. Then the door bursts open and he enters like a storm, dragging the sour stink of liquor behind him, his presence filling the room and turning the pastel air brittle. In his hands is a cake gleaming.
7:47It's pink frosting too smooth, like plastic dipped in sugar, like something that belongs on a screen, too perfect to hold. Every noun having an adjective attached, certain kind of repetitive syntactical blocks that appear. So there's all of that. There is all of that. but that on its own you know AI chatbots write like that because humans did once you know like it's an aggregate of all um human writing it can get its hands on so I would kind of think that that on its own isn't quite enough that's what makes it so difficult right because you can find that in human writing too but it is and this will sound wishy-washy but I'm I I think people will know what I mean if they've read large chunks of AI writing, is that there's something that happens when you read AI-generated text over a long, like something as long as a novel, is that there is just this like spidey sense you get about it of flatness.
8:46It's just, it's very emotionally one note. There's very little variety in the texture of the prose. And yeah, it's a feeling. It's a feeling where you sense literally a lack of mind behind the text.
9:00Noel King:So you've been clear that we don't actually know the truth of what happened here. We have an author saying an editor must have inserted AI later on. You understand publishing a lot more than I do. Do you find it believable that that could have happened? In a way, yes, because it's possible. I don't think it's possible for it to have happened without her noticing. I think she must have known about it if, if, if, if a third party did. I mean, it sounds, obviously, I have no idea. It sounds unlikely that it was done by a third party. If I had to nail my colours to the mast, I would think that probably it was her.
9:36Obviously, I have no idea. But that seems like Occam's razor, most likely. But then it ought to have been picked up when it went through Hachette. You know, the problem with self-publishing to traditional or publishing pipeline is that because the book is complete and also because the book has a proven audience already, editors are therefore maybe reasonably, they take the book in and they think, well, not much needs to be done with this because it's a complete work already. You know, it might not be as rigorous an editing process as it would be if just an unpublished first draft came in on an editor's desk.
10:09Another is that editors, unfortunately for everybody, including themselves to have less and less time for editing like in the major publishing houses people talk about this all the time within the industry that so much more of their work in recent years has been given over to you know stuff that needs doing like campaign work or liaising with authors but it is leaving less and less time for them to do the work of editing because the publishing houses don't have enough staff and all these kind of bigger picture issues
10:42Noel King:So where we've ended up here is that readers feel betrayed. A writer's career is more or less destroyed. Hachette looks kind of foolish. They do. It makes you wonder whether anyone is talking about guardrails in place to keep this from happening again, or whether at the moment it's just like, we've got to wait until the sleuths on Reddit figure out that there's a soullessness to this thing and start asking questions. I mean, I think everybody is talking about what those guardrails could or should be. It's a very hot topic, I think, among editors all over the place. But it's a really tricky one because AI detection software is the sort of first thing that you might think, okay, well, everything should get run through one of those.
11:25Those, when you dig into them, pretty unreliable. Oh. Yeah, they're not as good as one might hope. I mean, they'll pick up extremely obvious instances of AI, but they may not pick up, they just may not pick up everything. And then you've got this difficult thing. Let's say you have a manuscript from an author and you put that through an AI detection software and it comes up like 70 % likely AI. you know depends who that is if this is if this is an author that you are very keen not to you know it's relational stuff the publishing industry you don't want to accuse someone of doing something that at the moment anyway in the current cultural climate is kind of insulting like did you actually not write this book yeah did you get did you cheat you know um that's a really difficult conversation to have unless you're completely sure and it's really difficult to be completely sure.
12:18The thing about AI writing is that it cannot be original in flavour. That's the very nature of it. It's an averaging out of everything it can get its hands on. I don't know. I mean, maybe this position will sort of die out, which is depressing to me. But I feel like I would rather read a bad book written by a human than a quote unquote good one written by AI because at least, you know, to me, to me, what is valuable about a book is, is there, is it as a piece of art and to know that somebody had like something they wanted to say and maybe they will succeed at that and maybe they will fail. But that that's, that's the kind of transaction that you enter into with an author.
13:04It's like, okay, here's, here's what I tried to do, like meet me there with it. But if it's just something that has been churned out by a machine with like no love for the game, no craft, no intention, then I have no interest in reading that.
13:23Noel King:Imogen West Knight is a writer. Coming up, could your closest friends tell the difference between something you wrote and something AI wrote? An author experiments.
13:39Support for the show comes from Chime. Banking can sometimes feel like you're playing a game of dodgeball. Love dodgeball. Doing everything in your power to not get hit with overdraft fees, minimum balance requirements, and monthly fees. Don't love that. Chime wants to help. They say that they're approaching things differently and changing the way people bank. Chime says they are offering you the most rewarding fee-free banking, banking built for you. Why pay to get your own money? They're not looking to charge you overdraft and or monthly fees, and they have thousands of fee-free ATMs for you to use built for you, not the 1%.
14:11Take that 1%. Chime members can even benefit from up to$1 ,150 in annual rewards fee free plus direct deposit unlocks the most rewarding way to bank at Chime. Chime says they're not just smarter banking, they're the most rewarding way to bank. You can join the millions who are already banking fee free today. It just takes a few minutes to sign up. Head to chime.com slash explained. That's chime.com slash explained. Chime is a financial technology company, not a bank. Banking services and the secured Chime Visa credit card are provided by the Bancorp Bank N.A. or Stride Bank N.A. Optional services and products may have fees or charges.
14:42See chime.com slash fees info. Terms apply. Limited time only. Must open the new account and complete qualifying activities to earn rewards. Advertised annual percentage yield with chime plus status only. Otherwise, 1 % APY applies. No minimum balance required. Chime card on-time payment history may have a positive impact on your credit score. Results may vary. See chime.com for details on applicable terms.
15:00Noel King:Support for Today Explained comes from talkiatry. If you're dealing with depression or anxiety, you do not want to be stuck on a wait list to see a psychiatrist. You probably don't want to jump between online mental health platforms while trying to get medication support either, says Talkiatry. Talkiatry says they can help. It's virtual psychiatry designed to fit your life and your insurance. Talkiatry says they're a 100 % online psychiatry practice offering comprehensive evaluations, diagnoses, and ongoing medication management for conditions, including ADHD, anxiety, depression, bipolar, OCD, PTSD, insomnia, so much more.
15:33Noel King:Unlike therapy-only apps, TalkHiatry focuses on psychiatry, meaning you are seeing a medical provider who can diagnose conditions and prescribe medication if and when appropriate. They say you'll meet with an experienced licensed psychiatrist who takes time to understand what you're dealing with, creates a personalized treatment plan, and can prescribe medication when it makes sense for you. And they say you can book your first appointment in days, not months. You can head to TalkHiatry.com slash explained and complete the short assessment to get matched with an in-network psychiatrist in just a few minutes.
16:00Noel King:That's TalkHiatry.com slash explained to get matched in minutes.
16:08Noel King:Support for Today Explained comes from ServiceNow. ServiceNow says AI was supposed to handle the parts of the job that you hate. Instead, it just describes them, suggests what to do about them, and then leaves you to do it. What? That's not help. That's homework. ServiceNow's AI specialists are different, they say. They're not a tool. Think of them as digital teammates who actually do the work from start to finish. Cases get resolved. Requests get processed. Loops get closed. And most importantly, they say no extra work for you, because when you can truly delegate to AI, you can get back to the work only you can do.
16:38Noel King:The work that requires a person with ideas and judgment and, you know, a pulse. To learn how to put AI to work for people, visit ServiceNow.com.
17:06Noel King:Nothing we love more at Today Explained than a person running an experiment on herself. Wahine Vara, writer, journalist, author of Searches, out in paperback now. Tell me everything. So I think there's this, I don't know, misconception among writers, among readers, that there is a certain kind of way that AI generates language, and it's super different from the way writers do. And it was seeming to me from some research that I've read and from my own experiences that that was probably not true. So I wanted to do a kind of strange and dramatic thing, which was to see if people who knew my work really well could distinguish between my work and AI-generated imitations of my work that were created in a specific way that made them similar to my work.
17:59so there's a researcher named tuhan chakrabarty whose work i've covered before and he had already conducted this experiment he and colleagues basically trained ai models on the work of established accomplished writers um and what that means is he basically got the ai model to generate language that looked a lot like language from those authors And then he had readers who were graduate writing students read those passages generated by AI and also read imitations by fellow graduate writing students and say which one they liked better. And they tended to like the ones by the AI models more than the ones by actual human beings.
18:47And so I had him do the same thing with my work, but a twist on it. I had him train an AI model on my previous books, my three previous books on pieces of journalism I've written. And then I had him get his AI model to generate passages sounding like something from a forthcoming novel that I haven't published yet or shared with anyone. And then I put that alongside passages that I had written. I sent those to people who know my work really well. I'm talking about like my best friends since I was 13, writer friends who I've known since I was 19, 20 years old. And I asked if they could tell the difference.
19:32and none of them could.
19:39Noel King:All right. So the people who know you best in the world don't know you that well, apparently, or AI is exceptionally good at what it is doing. Give me some examples of what happened here. Can you read me something maybe that you wrote and then something that the AI wrote? And let's see if, I just met you, but let's see if I can tell any differences. It's funny, because I can't remember now which ones are mine and which ones are the AI. Gaia said it seemed to her that we'd been on similar trajectories. We'd both spent many years creating something that we cared deeply about. I with my journalism, she with her startup, and then gone on to focus on empowering others to do the same.
20:20She said she'd been surprised to find that mentoring other founders was even more meaningful than running her own startup. In business terms, the ROI was higher, if you are We're willing to count fulfillment as a return.
20:32Noel King:That's nice. I like that. Yeah, that was, I would say as writing, that was nice. Beginning, middle, end, lands on a point. I enjoyed it. That one was actually AI. No! Damn! AI, AI, you landed in such a nice spot. Okay, okay, girl. Read me something that you wrote, please. Okay. I guess, yeah, now we have a spoiler that I'm going to read you something from me. Okay. I'd like to argue that we write because we feel compelled to, no matter whether anyone will read them. But is that true? When I was younger, I used to keep a journal for myself. I didn't want anyone else to ever read it, which meant I didn't need to describe the people and places I was writing about or explain why they mattered.
21:23When my mom did read my journal in the ninth grade, I considered it the biggest betrayal I'd ever experienced. But the saving grace was knowing that she could not have possibly understood most of what I was writing about. I had an audience of one myself. Much better.
21:37Noel King:I liked the AI, but that was obviously much. I don't know, I set you up to say that. No, no, actually, you didn't. You didn't. I would be very honest, and I did sort of want to curveball you, but that was very pretty. Do me a favor. read the first two sentences of what you wrote one more time for me. I'd like to argue that we write because we feel compelled to, no matter whether anyone will read them. But is that true? Wait, we write, what is the them referring to? It's an error. It's a grammatical error on my part. Look at me. Okay, so. And good job catching it because a lot of people assumed that one was AI.
Read the full transcript
22:17And I think the best indication that it was actually me is that there is that grammatical error because AI wouldn't have made a grammatical error like that.
22:26Noel King:This is the thing that I would like us to talk about. AI does not make mistakes. And in the first half of the show, our guest, also a writer, described AI as kind of soulless. And I think that was part of what she was pointing to. What you read me by the AI wasn't bad. It sort of seemed like something that I might read in like lean in. Right. You enjoy mentoring other people. I'm going to hear why. So question for you. When all this was said and done, people could not tell what was you. People who know you well couldn't tell what was you and what was AI. What did you feel about that? Did you feel threatened?
23:07Noel King:Did you feel suspicious of your friends and family? Um, you know, I was of two minds because on the one hand, I didn't feel threatened, but I found myself questioning my own assumption about myself, which is that I identify as a writer who is very invested in originality, who really wants every new book to be completely different from the previous books. And so the fact that this AI was trained on my previous books and could predict the style of the writing in the new book suggested that I wasn't as original as I thought, that my new book wasn't as different from the previous books as I thought.
23:56And at the same time, on the other hand, I actually felt vindicated because I disagree with the other author who is your previous guest about the soullessness of AI-generated text. I don't think that AI-generated text is by definition easily distinguishable from human text because of a kind of soullessness inherent in the text.
24:21Noel King:Okay, but can most readers tell that something is AI versus something written by a human? It seems like they can't. And I can't myself. And this actually gets back to what we were discussing earlier about the question of whether AI-generated text is convincing or soulless. And I think the reason a lot of people assume AI writing is going to sound soulless is that AI companies in their most recent versions of their products have created these products that are specifically designed to sound a certain way, like a certain kind of corporate customer service speak kind of way. And so people think that's just what AI sounds like, right?
25:10Like that's somehow inherently the way AI sounds, but it's not true. AI can sound any number of ways. So it's technically very easy, actually, to build an AI, to train an AI model that sounds human-like, even literary. The reason we're not that familiar with it is that that's not what the products look like currently.
25:33Noel King:So ultimately, do you think AI is going to end up changing our relationship to literature? Do you think everybody who reads is going to be as skeptical and skeeved out as you an IR. Well, research shows not only that in some cases people prefer AI generated text to human generated text, but also that if they're told that a piece of text is AI generated, they become uninterested in it. And so it seems clear that people in general, the reading public, does not want to read text generated by AI if they know that it's generated by AI. You know, I think we focus a lot on this human technology binary, on like, oh, it's weird if a machine creates the language.
26:25But I think a big part of it is that we want to be communicating with one another. We don't want to be receiving our art from enormous tech companies that have a lot of wealth and have a lot of power and want to control us, right? So for me, it's really about do we want to communicate with other people or do we want to receive text from enormous technology companies?
26:56Noel King:Wahini Vara, writer, journalist, contributor to Businessweek and The Atlantic and The New Yorker. Her book Searches is out in paperback. Kelly Wessinger produced today's show and is the voice of shy girl. Amina El-Sadi edited. Gabriel Donatov, Check the Facts, and David Tattashore and Bridger Dunagan, Engineer. I'm Noelle King. It's Today Explained.
From the publisher
An author gets her book pulled after accusations that it was written with AI, but it might not always be so easy to catch AI writing.
This episode was produced by Kelli Wessinger, edited by Amina Al-Sadi, fact-checked by Gabriel Dunatov, engineered by David Tatasciore and Bridger Dunnagan, and hosted by Noel King.
Photo by Planet One Images/UCG/Universal Images Group via Getty Images.
Listen to Today, Explained ad-free by becoming a Vox Member: vox.com/members. New Vox members get $20 off their membership right now. Transcript at vox.com/today-explained-podcast.
Learn more about your ad choices. Visit podcastchoices.com/adchoices




