A $2 million offer for a debut novel canceled last week due to suspicions—no proof has been given so far—of AI use.
This is one of the decade’s biggest publishing stories. Those who don’t know the book world believe advances of this size are common. In fact, they are extremely rare for non-celebrity debut fiction. The best writer in your college class… well, she probably became an actuary, or applied to medical school, as soon as she learned about the query system… but if she, for some bizarre reason, ever did try to get a book published, she’d be lucky to get $20,000 in today’s climate. The truth is that most people will never accrue the social capital necessary to get read by a single literary agent. Manufacturing the concurrent interest—a 14-way bidding war is unusual, to say the least—that produced that offer requires luck and skill both in extraordinary amounts. If the novel was indeed written by AI, the debut-fiction game has been solved by a GPU cluster—this is the AlphaGo moment. If no AI was used, this is a literary injustice that could end careers and will be remembered for decades.
The evidence behind the accusation was shockingly thin, to the point that the agent who interfered with the offer can be convicted of recklessness even if proven correct. Negative parallelism (“it’s not X, it’s Y”) and wooden prose were cited, but overuse of rhetorical devices has been a hallmark of the commercial bestseller for at least a century, and so has bland writing. Copywriters at ad agencies have been using correctio, parataxis, and tricolon (“rule of three”)—often very poorly—for decades. Linguists study these devices to understand how pragmatics and grammar evolve. Most elite writers have probably read one or two craft books on these techniques, but use them sparingly. When negative parallelism works, I’ll use it, and I don’t give a shit what AI does, because I don’t give a shit what inferior writers do, and AI is an inferior writer. (Syllogism, motherfucker.)
I don’t worry about AI writing or AI accusations. Anyone who thinks AI can currently write as well as I do has no sense of taste and can be ignored for being a fucking idiot. If, in the future, AI develops the ability to write at my level or above it, then we have bigger problems than the death of the fucking novel.
I predicted in January 2023 that the commercial novel would fall. It doesn’t much matter if this particular novel was AI-generated; the fact that publishers believe they’ve made a $2-million offer for chatbot slop means that they have lost the ability to tell the difference between mindless entertainment made by mindless machines and traditional publishing’s product, which is mindless entertainment mindfully optimized by humans for pitchability and corporate approval.
I wrote a satirical essay, “How to Make AI Write a Bestseller”, in May 2025. It’s already outdated. It describes the drudgery it would take to produce commercially competent AI slop at book length. The conclusion is that it would take as much effort as writing a real novel, and be unpleasant. Claude Code barely existed. These days, the method in that essay could be performed using sixty hours or less of human labor. Would such an approach produce a truly good novel, the kind of book people remember fifty years later? No. Would it guarantee bestselling results? Nothing in text guarantees bestselling results, so also no. Would it produce a novel that, if given the editorial and marketing investment afforded for a lead title, stands as good a chance of selling a million copies as any novel produced by any human, including myself? Yes. It would not be work that wins the respect of those who understand text at its highest levels—literary writers, lifelong serious readers—but none of these people’s approval is necessary for the next Fifty Shades of Grey.
♪ All debts are paid… at the slopera tonight. ♪
I hate saying this. Commercial storytelling is probably dead. It’s bad news for everyone, because even the snobbiest literary authors rely indirectly on commercial fiction. Elite novelists’ tenure lines are funded by tuition paid by future Hollywood screenwriters. The fact that summer blockbusters can be produced with minimal human input—trust me, I’m not the only one out there who’s smart enough to make a machine do it—is probably good for nobody, save Sam Altman.
Of course, markets move. Give me a year and a $5,000 token budget, and I can produce a commercial novel—fuck it, twenty of them—that, with the right kind of social access, could have been a seven-figure lead title in 2024. In 2027? No one knows. The seven-figure advance might just die. The poor door called the query system always offered abysmal odds of resulting in a book deal worth taking, but we can expect traditional publishing to become even more trust-based and exclusionary, because AI is the perfect excuse for publishing to legitimize the culture of favoritism and nepotism it has always had and always wanted. It used to be said in publishing that anything good in the slush pile was probably plagiarized, but this accusation could be refuted. AI abuse can’t be.
One of the foremost topics of discussion among publishing executives right now is whether to use AI for continuity after their bestselling franchise authors die, but when it comes to debuts and new releases, they are far more likely to retain human authors for an ugly reason. Traditionally published authors learn quickly that remaining publishable requires them to meet sales targets that, without a lead-title push, are impossible to achieve without excessive investments of time and money into their own publicity. In this context, it remains better, from publishing’s standpoint, to offer $5,000 advances to people whose lack of social access prevents them from securing better deals. In this sense, commercial storytelling is already dead—the overwhelming majority of published authors, enticed by agents’ and editors’ promises of future literary careers, write their stories for free and accept less than minimum wage for freelance growth hacking.
Literary fiction’s fate is even more grim. Publishers closed the doors to new authors of epic fantasy in the late 2010s because they grew tired of authors in other genres using fantasy comps to negotiate for higher word counts, and because they were sick of agents using on-screen potential as an excuse to argue for decent advances. There are a number of reasons why they want to cut literary fiction, but one of the biggest is that they are tired of “upmarket” positioning being used to justify an author’s refusal to turn characters into vampires, werewolves, or mindflayers if an editor demands it. Publishing is unlikely to fire every single one of its kept literary novelists, but no one who hasn’t already gotten in will get in, because human leverage, in a world where LLMs can do basically any verbal task at a mediocre level, has reached zero.
Commercial fiction—derivative, optimized for entertainment value, easy to polish—will continue to have publishers’ support, although it will be increasingly overseen by “writers” who will work on twenty titles at a time, making sure the robots don’t do anything terrible like sneak an axe murder into a children’s book. Literary fiction will still exist, because people will want to read and write it, but I have no idea how we’ll get it paid for. The market-clearing price of an unknown author’s books remains negative—it costs more to find readers than one can charge them—and, while institutional support can remove expenses, the moral and intellectual costs of retaining said support tend to eliminate the precision-of-thought that serious literature requires.
I know that people will tell stories in 2040. It’s ironically one of the most honest things we do. I don’t know if anyone will make money by telling them. That might be truly over. And this might be a good thing for society. Consider that millions of Americans believe they’re going to write a bestselling novel someday. This dream is one of the factors that keeps people believing in capitalism. They vote against their own economic interests because they might one day send a novel to an editor at a publishing house (they’ve never heard of literary agents, nor the query system) and cause a stranger’s heart to grow three sizes that day. They fantasize about seven-figure advances. They even believe text matters to publishing. So long as the possibility of “their novel” being written and picked up exists, they can sustain faith in a socioeconomic system in which they actually have no future. It might be better for all of us if this collective delusion goes away.
Personally, I have no interest in reading fiction written by artificial intelligence. There are thousands of excellent novels, written with real intentionality by real people, that I will die before I have the chance to read. Generative AI is often competent and it is often creative; I’ve never seen it be both. Still, we have strong evidence that it has mastered commerciality. These incursions will happen again and again.
AI Disclaimer: I used Claude Fable 5 for copyediting. I never mix generated language into my own prose, as I have too much pride to let AI do my writing for me. At the same time, understand that anyone too prideful to use AI in this way is also too proud to submit to traditional publishing’s query system and its general idiocy.
You can compete in writing. You cannot honestly compete under rules where the prize is awarded before the work is read.
You are not merely competing against other writers. You are competing against imported audiences, institutional prestige, professional networks, marketing machinery, inherited visibility, and now limitless synthetic production. Quality may matter after discovery, but quality cannot help when the work is never shown. (or read)
And the platforms benefit from calling this openness. They can say anyone may publish while accepting almost no responsibility for whether unknown writers can actually be found. Publishers increasingly want the same thing: not simply a manuscript, but a ready-made market. The author is expected to arrive already carrying the readers, so the institution can reduce its risk and collect the profit.
That is not a literary competition. It is an audience-acquisition competition disguised as merit.
When I saw the 2-million-dollar deal fall through, I was very excited to read your take. Thanks for these great thoughts!