What Defines Acceptable AI Use for Writers?

Recently BookBub surveyed 1,200 authors and found 45 percent are using generative AI in some form, whether for writing, marketing, research, or other tasks. Of those using AI, more than half are using it to help with writing.

When these figures were shared on social media, some questioned whether the figures could really be that high, and some were disturbed by the results. Many took the opportunity to condemn AI use or argue that any author using AI to write wasn’t, in fact, a real author.

I don’t have to tell this readership that AI use, across all creative fields, has become divisive. I’ve had people cancel their subscriptions because I cover AI like any other topic rather than discourage its use. Whether or not you decide to use it, it’s critical to understand how the industry is using it—especially publishers—and how perfectly legal AI licensing schemes are growing alongside the lawsuits. AI is part of everyone’s future and, as Marshall McLuhan once said, resenting a technology will not stop its progress.

So I am not surprised by the survey results, but most writers, out of fear of being attacked, don’t admit on social media or any public forum that they’re using AI. I do hear guilty confessions about AI use in private and offline conversations. Anyone who believes the majority of writers are resolutely against AI is shut off from reality. Writers are less afraid of AI than of being harshly judged by their peers.

But what does it mean, exactly, to use AI to write? It’s going to depend on the writer—every creative process is different—and the BookBub write-in survey responses reflect as much. What I find overlooked or misunderstood by the anti-AI crowd are the nuanced ways these tools get used. Calling it plagiarism is reductive and inaccurate. For example:

One journalist and author uses AI to conduct interviews with herself based on editorial feedback. Her answers, transcribed by AI, get transformed into a draft she can revise—with language clearly marked as originating from her or the AI.
Stephen Marche creatively combined the outputs of multiple models to publish the novella Death of an Author.
Vauhini Vara used AI to create falsehoods that she then wrote against. She says, “For me, somebody with 20 years of experience as a writer, who writes for a living and who knows a fair amount about technology … I was able to have this engagement with the large language model in which I responded in a way that, like, drew out my own perspective. …”
What about the university professor who asked his students to have a conversation with ChatGPT about the history of attention, edit down the text, and reflect on it? He says the results were the most profound of his teaching career (sub required).
I’ve used AI on drafts to identify lapses in logic, flag questionable claims, and speed up my research process. (That chart in my article about 8th Note Press? It was initially AI-drafted once I identified the titles to research.)

It’s telling to see writing and publishing people admit to using AI “just” for marketing and promotion—or say that it’s okay to let AI do the work that isn’t “creative.” (I would argue good marketing is in fact creative.) Or that it’s acceptable for other professions, such as science and medicine, to use AI, to fuel discoveries that will benefit society—but it’s not okay for creators to use it to fuel their own discoveries.

I see no shame in using AI as a brainstorming partner, beta reader, writing coach, or intern—and that’s not an exhaustive list. Moreover, I find it impossible to draw a hard line between “This AI use is okay for you and your work” and “This AI use crosses a line.” Nor do I see how strict enforcement or detection of AI writing will ever be possible; just take a look at these nightmare stories of false positives. When AI use is suspected today by publishers or agents, it’s typically because the writer failed to use it well. Dull, featureless, derivative work, whether or not it employs AI, gets rejected.

Obviously the tools can be used in harmful ways: All kinds of bad actors use AI to churn out copycat books, low-quality summarizations, and other garbage that clogs up submissions inboxes or the Amazon marketplace. (Or see this week’s example of a journalist who used AI to generate a bogus summer reading list rather than calling up a bookseller, librarian, or really any human being for recommendations.) But I believe we can separate out bad-faith, lazy, or fraudulent efforts from writers using AI as part of their bespoke creative process.

Another thing bothering me about response to this BookBub survey: Some people are quick to point out, “Well, these respondents using AI are self-published authors, not traditionally published authors.” That is partly true: 69 percent of the respondents are self-published, 6 percent are traditionally published, and 25 percent have done both. The majority are writing commercial or genre fiction.

But I do not like the subtext lurking underneath that statement: that self-published authors produce low-quality work or engage in questionable practices that aren’t befitting of “real” authors. Yes, successful self-published authors tend to be prolific and write works that might be more easily outlined and filled in with the help of AI. (Certainly before AI, and now, prolific authors may hire ghostwriters and assistants to help them write or finish books.) So I understand why self-published authors may be more likely to use these tools because of the pressures they’re under and the type of work they produce.

It’s a big mistake, though, to think traditionally published authors don’t experience similar pressures and temptations. Or that they’re not curious people who use these tools to support brainstorming, research, narrative development—any type of writing challenge. In fact, I find the more literary the author, the less worried they are about AI, legal issues aside. They may not see AI as meaningfully competitive or as a technology that could eliminate the meaning of what they do.

Meanwhile, I find commercial authors do worry what happens when writing becomes impossible to distinguish from human writing. The US Copyright Office recently posited that AI models may dilute the market for some types of work, using romance as a key example. There’s potential for profits in any genre where there’s high demand if AI can produce sufficient quality to please readers. Some believe commercial works are easier to mimic and its readers are less discerning. If that’s true (a very big if), how much will the origins of a story matter to the average reader? Or what if publishers use AI to continue producing work by big-name commercial authors even after they die, or use it to facilitate work on existing intellectual property? Is it good news for human authors when readers become suspicious of a debut author who comes out of nowhere with a commercial success and start asking: Is it AI generated?

We can get into a rich debate about all of this, but one can’t fault authors for fearing a dark outcome. And that’s the fear that fuels the fiercest condemnations. Yes, there’s anger over the original sin of these models training on copyrighted material without payment or permission, and there’s deep distrust and dislike for their billionaire owners. But will minds change when these tools are “clean”—only trained on licensed material? Unlikely.

I like what Stephen Marche wrote about AI: “Creative artificial intelligence provokes a strange mixture of contempt and dread. People say things such as ‘AI art is garbage’ and ‘It’s plagiarism,’ but also ‘AI art is going to destroy creativity itself.’ These reactions are contradictory, but nobody seems to notice. AI is the bogeyman in the shadows: The obscurity, more than anything the monster has actually perpetrated, is the source of loathing and despair.” Well worth reading his entire essay.

Bottom line: Just as I don’t like to see authors divide into tribes over how they choose to publish, I don’t like to see authors divide into tribes over AI use. I don’t see the productive end of policing how other people decide to write, edit, and publish their work. But for now, I do believe we’re in a “don’t ask, don’t tell” period. While there are copyright implications (that is, you’re not supposed to receive full copyright protection for purely AI-generated material), who exactly would bring court cases against such authors if they’re not transparent about AI use? Who will care or be harmed enough to invest in such a case? (Maybe publishers if they pay a high enough advance and can’t get it back.) While AI presents some of the most challenging questions I’ve seen in my lifetime as a creative professional, rather than fight to stop its use—an impossibility—we must confront head on what makes human writing distinct and valuable when machine-generated words can be summoned with a click.