Generative AI has now been in our lives long enough that no college senior graduating this spring has experienced a single year of college without it. Yet I don’t find that book publishers or writing awards have truly come to terms with the prevalence of this technology, how it affects behavior, and how it necessitates new processes and responsibility.
Any organization that wishes to prohibit AI use today must face an evolving conundrum: Writers are astute enough to claim, “This is my human work and you can’t prove otherwise,” regardless of whether they’ve used AI. Since the very institutions that prohibit AI use are also unlikely to implement AI detection software, they may unwittingly bring attention and investment to work that incorporates AI output. If writers don’t see methods of enforcement, they will use AI if it suits them—not because they’re bad people but because they’re human.
Recently, regional winners of the Commonwealth Short Story Prize were alleged to have been written with AI. I observed a publishing-industry professional comment on social media that the situation demonstrates that editors should be trained on what AI writing looks like. Another professional responded along the lines of “absolutely not”—they would not let such technology live rent-free in their head—yet offered no alternative solution.
This, I fear, is abdicating the responsibility that every publishing professional now has, which is not necessarily to learn AI but to effectively deal with its consequences in their work. It is possible to stand up against the misuse of AI, have concerns about its billionaire ownership, and consider its negative societal effects, all while also exploring responsible, informed use. While prohibiting AI might be appropriate for some, enforcing such a policy (if it is a real policy and not merely a wish) requires effort and resources that will have to be continually upgraded and assessed. How hard do you want to fight, and how much time and money do you want to commit? I can imagine a couple paths forward depending on the resources of the organization.
One path forward: Adopt a no-AI policy and determine the method of enforcement. This describes Microcosm Publishing, which laid out their policy earlier this year; they use AI detection software such as Pangram to aid in enforcement when working with authors. Some argue that detection software is insufficiently accurate, although I haven’t found anyone saying they rely solely on software to render a verdict. Rather, it is a tool that can raise a flag for further human review. Any organization could make it a condition of submission that the writer automatically grant permission for their work to be screened by AI detection software that doesn’t save the work or train on the work. Organizations that can’t abide use of such software (or don’t have the money) might need to employ staff or hire freelancers who understand AI well enough to detect its use or who research potential authors or prize-winners for signs they’re AI users. (Yes, literary organizations turn into the AI police. It’s not pretty.)
Another path forward: Require writers to certify their writing is human through a third-party service. I’m contacted by such services every month or so. They use varied methods of certification, some of which require special software that writers need to download and use, but I cannot envision this happening on a large scale. Other certification services, like Verify My Writing, are based on the same technology as AI detection software; Verify My Writing uses Pangram. Whatever the method, requiring writers to self-certify would likely put the cost burden on the writer. (Note that the Authors Guild’s Human Authored Certification relies on the honor system, so it’s not certifying anything other than that a verified human being has submitted the work and vowed “it’s my creation,” which doesn’t solve anything.)
What I think will happen over the long term: Anti-AI policies will fall away. First, I’m not confident AI detection software can keep up with the rate of AI advancement. Even if it can, there are increasingly finer shades of gray. Is 23 percent AI-assisted work, as determined by Pangram, a deal breaker? What percentage is an organization comfortable with? How do they decide? Why not just use standard editorial criteria for evaluation and reject what’s unacceptable as unacceptable and be ready to explain why? Institutions that don’t prohibit AI don’t have to police percentages or deal with suspicion, witch hunts, and scandals—or enforcement that’s likely to be contested.
That brings me to the IBPA Book Awards: Their guidelines do not prohibit AI-assisted work. One author who’s been transparent about his AI use, Luke Stoffel, just won the 2026 IBPA Award in the neurodivergent communities category for his memoir, How to Win One Million Dollars and BEEP Glitter! In an interview this week, Stoffel told me he’s dyslexic and has been using AI for three years to support his creative work. (He is also an active artist and creative director.) When he was young, he was never encouraged to write, and his teachers couldn’t read his garbled sentences; his sister would help edit his work before he submitted it to anyone.
When AI came on the scene, Stoffel had been working on his memoir since 2016 with a writing group and developmental editor. His initial instructions for ChatGPT were narrow: Fix the grammar, leave the voice alone. But he found the collaboration more generative than he expected. The AI helped him develop a three-dimensional story in ways he hadn’t been able to manage on his own, and he came to see his use of AI as thematically central to the book itself. The memoir’s final reveal is that he used advanced technology to compensate for the limitations of dyslexia and ADHD. Since then, he has written more works in collaboration with AI, ultimately using Claude to develop a science fiction novel exploring consciousness and his relationship with the technology.
“I would’ve never been a published writer without it,” Stoffel says. Plus, using AI didn’t mean less work for him. He still spent hours upon hours writing and editing. Kirkus Reviews said his memoir was written “with humor, panache, and heart.” Publishers Weekly BookLife scored it as 9.5 out of 10 and said it was powerful. Stoffel doesn’t think the book would’ve been better if he’d written it without AI, but he also sees the creative dangers: that it’s easy to become exhausted during the creative process, and “we start to accept what the machine is saying to us and we publish without doing due diligence.” Stoffel says he’s done far more creative work in the past two years than he ever could have imagined doing, across all the fields he participates in, because of AI. He’s not concerned about losing his paid work because, he says, “AI will never have aesthetics.”
Bottom line: Aggressive AI policing implies that writers are choosing convenience over craft, or that no defensible AI use exists in the writing profession. I’d prefer to treat writers as professionals who decide on their tools and creative workflow, then judge based on output, not process. Of course, individuals with zero tolerance for AI typically have deeply held moral or ethical objections to the technology, but as I see it, many commercial publishers or institutions with diverse stakeholders are not proactive but reactive, mainly trying to stem online backlash, as was the case in the SFWA community regarding the Nebula Awards. The people speaking against AI are loud, but they’re rarely the ones who have to find a defensible and effective way to police AI use.
Stoffel said the IBPA judges knew he used AI but awarded him anyway. “It was a risk. The IBPA is involved whether they know it or not.” Will a scandal follow the IBPA as Stoffel’s award story spreads? I expect to hear a chorus of people object to AI-assisted work qualifying for and winning a major book award. At the same time, I find unwillingness to accommodate people like Stoffel—who are using the tools to execute their creative vision, not take shortcuts—unenviable. AI output varies widely with human input; it guarantees neither excellence nor atrociousness. What is the goal of excluding AI-assisted work in the long run if the author succeeds in the execution of their vision?
While there are a lot of jokes circulating about the bad writing among the regional winners of the Commonwealth Short Story Prize, to me, this isn’t a problem with AI. Rather, it reflects poorly on the judges’ taste.
Copyeditor Nicole Klungle assisted me greatly with insights into ableism, disability, and AI, and her questions on the topic contributed to my interview preparation for Stoffel as well as to the final bottom line.
Jane Friedman has spent her entire career working in the publishing industry, with a focus on business reporting and author education. Established in 2015, her newsletter The Bottom Line provides nuanced market intelligence to thousands of authors and industry professionals; in 2023, she was named Publishing Commentator of the Year by Digital Book World.
Jane’s expertise regularly features in major media outlets such as The New York Times, The Atlantic, NPR, The Today Show, Wired, The Guardian, Fox News, and BBC. Her book, The Business of Being a Writer, Second Edition (The University of Chicago Press), is used as a classroom text by many writing and publishing degree programs. She reaches thousands through speaking engagements and workshops at diverse venues worldwide, including NYU’s Advanced Publishing Institute, Frankfurt Book Fair, and numerous MFA programs.
Jane wrote with respect to AI use, “I’d prefer to treat writers as professionals who decide on their tools and creative workflow, then judge based on output, not process.” While I respect Jane immensely, I beg objection. If this same approach were used in the Olympics, athletes using performance-enhancing drugs could argue these drugs were merely “tools” for maximizing their “output.” However difficult (or impossible) the policing of AI use in writing will become, a line has to be drawn between what should be considered literature or a cheat.
Ghostwriting, editorial collaboration, research assistants, dictation, and even spell-check have always aided authors’ output, and (I’d argue) maximized it. Also, writing and publishing isn’t in my view a zero-sum game. Does a writer who uses AI cheat another writer? Maybe there is an argument that it does, but I think the writer who uses (or leans on) AI is mostly at risk of cheating themselves. Is the writer cheating the reader if use is known/disclosed?
Thought-provoking points, Jane. However, ghostwriting, editorial collaboration and research assistants are all driven by humans, not machines. Dictation merely transcribes what a human (hopefully) has written or spoken; like spell-check (which is corrective in a purely technical sense), it is not generative. Nobody will dispute the value of Lennon collaborating with McCartney and vice versa. But if McCartney “wrote” an AI-generated song, would it have the same intrinsic value?
Probably not, but that’s a different question/issue than whether AI use is cheating.
Great read about AI. Thanks, Jane. AI is here to stay, so people will, in their own ways, figure out how to deal with it. I read daily, including many articles on Substack, and my default is to assess the quality of the article and its writing without really thinking about where it came from. I just want to read useful and entertaining material.
However, I must confess that AI has opened up a new world for me. I am currently helping a 92-year-old retired doctor friend write his autobiography, along with a second book. He is a self-published author, but due to two strokes and eye issues, he can no longer read. With him on the east court and me on the west, we meet by phone for 75 minutes daily, and I record our conversations using Plaud AI, which then provides excellent transcription. Next, I use Claude to organize the massive amount of information, not by rewriting it or changing his voice, but by keeping our notes and structuring them in a way I can use. Then I read it all back to him, and he offers changes which are again recorded by Plaude with Claude serving as our assistant editor. We have now built 39 chapters with 68,000 words over 95 sessions. Without AI, this would be impossible. I think it has given my friend a new reason to live — he can still be productive! I am so thankful for this technology.
What a great use of AI! And it shows how using AI in this way enhances human connection.
For me, the question is: is it ethical to allow the use of generative AI in awards/competitions (given that the LLMs are using stolen material scraped off the internet without the permission of and without recompense for any of the authors) before there are systems in place to allow opt in/opt out and to compensate authors fairly?
I think it’s fair to say authors/publishers are being compensated and will be compensated further, due to countless licensing agreements with publishers, and the $1.5 billion Anthropic settlement.
Um, no, sorry. You’re wrong about that. Four of my novels were scraped by Anthropic but the cut-off date was too late for me to be included.
A British novelist, I’ve always paid the Library of Congress to hold my copyright, so in THAT respect I’m eligible for compensation, but not time-wise, regardless of my work being pirated and scraped.
Very many other British writers aren’t eligible because it never occurred to them that their UK publishers wouldn’t register their works with the Library of Congress and they never saw the necessity to do so independently. And their novels were pirated/scraped just as much as mine!
So, no, SOME authors will eventually get SOME compensation. Others – equally worthy of it – will get nothing.
Alice, I’m sorry your work isn’t eligible in the Anthropic case. Regardless, there are ongoing licensing schemes that are paying authors and publishers alike for AI training, as well as more than 100 other lawsuits in progress. We probably have another decade of litigation, if not more, involving copyright claims.
As a new writer who has finished her first manuscript, it is difficult to understand where the lines are drawn. As I revise, if I ask AI to look at a chapter and it tells me there is an error in repetition and flow, and I’m the one to fix it, it’s just another tool, like Grammarly. Today on Substack, an author put her work through Pangram and the software said it was 70% AI. This is alarming because the author was conducting an experiment on words that were only written by herself. These detection tools are faulty and the lines are quickly blurring.
Side note, I stumbled upon this article through Threads and coincidentally will see you at the Stockholm Writer’s Conference. Should be an interesting discussion on AI!
Thank you, Lorraine – see you soon!
Your question/suggestion “Why not just use standard editorial criteria for evaluation and reject what’s unacceptable as unacceptable and be ready to explain why?” makes great sense to me. In my work as an editor I read a LOT of human-generated content–which I know was such because there was no AI then–that sounded stilted. I’ve evaluated stories for anthologies with tired, cliche-filled plots. AI doesn’t have a lock on these things, and it will only get better. Much of our discomfort comes from knowing that a non-human was involved. And writers aren’t the only creative people facing this; musicians, visual artists, and even game designers are in the same situation.
Well said, Jane.
Let everything, AI is just a tool. It’s how we use the tool that makes the difference.
I object to the people using AI as a shortcut, not bothering to edit or scrutinize the output. But if writers are using it to spark ideas and to allow them to dig deeper into the material, then they weren’t using AI to “generate” their story. They were using it like a critique partner or prompt generator. It’s different.
I 100% agree that it all needs to come down to the end product. If the end product is found to be substandard but still published, then that’s a failing of the people who chose to let that substandard work pass. However, if the end product sparkles and shines, then why shouldn’t it be out in the world?