The main floor of Capital Books in downtown Sacramento. (Photo by Eric Schucht)

AI Is Changing the Book Business. The Literary Community Is Pushing Back

Capital Region writers, booksellers and publishers share concerns over generative AI books

Back Web Only Nov 18, 2026 By Eric Schucht

Approximately 12,000 different titles are for sale on the main floor of Capital Books in downtown Sacramento. Of those, bookstore co-owner Ross Rojek said the number explicitly written by generative artificial intelligence: “Zero, but written by AI with a human name slapped on it, I don’t know.”

Large Language Models like ChatGPT and Claude are increasingly used to churn out machine-written novels. Some have embraced the technology, while those in the publishing industry, like Rojek, avoid it like the plague. He refuses to stock any generative AI titles, believing they won’t sell. “People want to buy books by people,” he says. AI book covers and artwork are easy for Rojek to spot, but he increasingly doubts what is or isn’t machine-written text.

“It’s not going to get any better,” Rojek says. “It’s going to get harder and harder to detect, and I don’t know that there’s a solution to it other than human curation and customer decision.”

Publishers and authors in the Capital Region have pushed back against the recent proliferation of AI books, refusing to touch the technology or work with those who do.

‘I just don’t think that machines can feel’

Julia Park Tracey, of Grass Valley, is the executive editor at Sibylline Press and handles acquisitions. She screens all submissions sent to the publisher, which exclusively works with women authors over age 50, and rejects manuscripts suspected of AI text, believing them low quality and dishonest.

“I just don’t think that machines can feel,” Park Tracey says. “They can fake it, but we want authenticity. We want vulnerability, and how can a machine be vulnerable?”

Sibylline Press hasn’t received many manuscripts suspected of AI text, but a recent incident proved to be a grey area. An aspiring author admitted during the signing process that she used AI to rewrite and copyedit her book. The company asked her to rewrite it herself. If someone has a story to tell but lacks experience, Park Tracey encourages them to hire a ghostwriter.

“If you don’t have the skills to write a book, work with a human,” Park Tracey says. “But feeding it into AI, it just sets off red lights all down the production path because I don’t want to waste my time reading a book that’s not authentic.”

Sibylline Press publisher Vicki DeArmon, of San Diego, said that using AI-generated text in a book “is a mark against quality, but it’s also a mark against promotability.” Publicizing who wrote it, how and why are key to marketing. These traits are absent in a machine-written novel, making it a tough sell. “It’s not like any book rises to the top without a parallel construction of who the author is,” she says.

DeArmon thinks generative AI could write a passable plot-heavy novel. For the more complex books Sibylline Press specializes in, like memoirs, an author’s tone, voice and uniqueness are vital for marketability, and AI text is too generic to meet audience expectations the publisher has cultivated.

“I’m not sure that there isn’t a market for it, but it doesn’t necessarily have the sophistication that we require,” DeArmon says. “But, you know, maybe those readers don’t require that sophistication.”

‘They’re getting away with it’

To develop LLMs, some tech companies have pirated books for use as training data without permission or compensation of authors, which has soured a lot in the publishing industry on the technology. “Many of us are already burned, mad, exhausted and feel ripped off by generative AI because we’ve had our hard work that we worked on for years stolen for training purposes,” Park Tracey says.

This has resulted in several class action lawsuits. Facebook owner Meta was sued for using pirated books from the file-sharing website Library Genesis as training data for LLaMA 3. Two of those books were authored by Park Tracey. The plaintiffs in Kadrey v. Meta Platforms, Inc. alleged copyright infringement, but in June 2025, a judge ruled it fair use. “Those of us whose work was used don’t call it fair,” Park Tracey says.

The authors lost their training-related copyright claim. However, a similar lawsuit, Bartz v. Anthropic, resulted in a $1.5 billion settlement. Anthropic used 7 million pirated books as training data for Claude. While the judge also found this fair use, they also ruled that downloading and keeping the pirated books was not, which led the tech company to agree to pay around $3,000 for half a million qualifying books.

Four of those titles were authored by Kim Culbertson, a recently retired high school teacher from Nevada City. As to the amount awarded, she thinks it’s too low. “I don’t feel that’s fair,” she says. The settlement money is split between the book’s author and publisher, so she’ll only net around $6,000 in total. Regardless of the amount, she sees the ruling as a win for writers.

“The fact that the judge cited on behalf of authors and on authorship and creativity and intellectual property, I mean, that’s a good sign,” Culbertson says. “That was validating to hear, ‘yes, they stole this from you.’ Did they make billions of dollars anyway? Yes.”

Elk Grove author Mark Steensland had one book qualify for the settlement and will split the award with his publisher. While any money is better than no money, if Anthropic had offered him that deal, “I would have said no.” Steensland feels like a lump sum is a means to buy off writers, and offering residuals, a percentage of any money made off their work, would have been more equitable.

“If you’re going to be having this kind of profit, then I should be sharing in that also, since this is my work that is allowing you to do this,” Steensland says. “So just in principle the arrangement is unfair, and it’s kind of like they’re getting away with it.”

Michele Drier, a retired journalist turned author from Sacramento, had three books qualify, all self-published, meaning she’ll keep the full settlement. As for the dollar amount, she said, “it’s a reasonably fair amount,” with the caveat being she isn’t an expert on intellectual property and trusts the deal the lawyers reached was just.

As for her book being used as training data without permission: “I’m really upset about it,” Drier says. Pirating her work, “I don’t like that, but it’s somewhat understandable because at least you stole it because you want to read my book.” But using her work for generative AI training data, in her view, is inexcusable. “Because then you’re making money off of it.”

‘It’s really dishonest’

Some aspiring authors have used generative AI to create machine-written books and fraudulently passed them off as original work. Others have used the tech to copyedit their writing, permissible to some, but unappealing to Culbertson. She enjoys spending time tinkering around with a sentence until it’s just right. “Efficiency is never going to be my drug of choice. I don’t think being faster is what makes something better,” she says.

For Culbertson, writing involves making creative choices that she believes reflect a person’s thought process and personality. She worries that using generative AI to rewrite text will cause authors to lose their individuality.

“Writing is about messing around with language and sound, and there’s a rhythm in it, and it’s beautiful,” Culbertson says. “And I am afraid that too much AI-generated text is going to all start sounding the same. It’s like a homogenization of language in a way that I’m not interested in.”

Steensland said the prevalence of low-quality generative AI books damages public perception of the art form. He refuses to use the technology, believing it utilizes stolen work and wastes electricity and water. He also points to other flaws, such as how chatbots have created misinformation due to “hallucinations” and worsened some people’s mental health to the point of suicide.

“We can clearly see the dangers and drawbacks of the technology as a whole, and that makes us responsible for it,” Steensland says, “and I’m not going to participate in that in any way if I can.

“I smoked cigarettes for years, and I thought it tasted great and it smelled great, and I looked cool as hell when I was smoking. Cigarettes killed my dad, and I wasn’t acknowledging objective reality. That’s something we need to do.”

Drier also won’t use generative AI for writing. “It’s really dishonest to put out machine-generated material that has no heart or soul to it,” she says.She thinks generative AI could be used to create passable “formulaic fiction,” a story that follows popular tropes and patterns, but worries these low-effort AI books oversaturate the market and “absolutely undercut legitimate authors.”

In 2023, Amazon’s self-publishing platform, Kindle Direct Publishing, announced a policy change to deter AI books from flooding the marketplace. Authors were allowed only three title uploads a day, or 21 a week. Two years later, that limit was reduced to 10 titles a week, and reduced further in September to two a week.

Recently Drier’s sales on the platform have declined. She doesn’t know why but speculates the abundance of AI books makes it harder for potential customers to come across her work. She hopes the recent policy change will help, but at the end of the day, it will be up to readers to decide whether or not AI books have a place in the market.

“I wish that there were some way that readers could be discerning enough that they would stop buying those books,” Drier says. “That’s the only thing I can think of that’s going to stop this.”

–

Never miss a story: Get Comstock’s weekly newsletter delivered to your inbox.

Recommended For You