It’s Getting Harder to Spot AI in Contemporary Publishing. And That’s Very, Very Bad.

Lately there has been a lot of hand-wringing, and rightly so, about if and how the publishing industry will deal with AI in the wake of the cancelation of the first major book deal due to suspected AI usage.

There is no easy solution to AI detection for many reasons, partly because large language models are imperfect replications of prose written by actual authors in the first place, and partly because large language models are improving (if “improving” is the word to use for a product built off the backs of a wide variety of human writers) all the time. The word salads that we might identify as AI today may not be the kind of machine-made writing that we will see tomorrow. Not to mention that some authors happen to naturally write word salads as a matter of course.

This is just the latest example of how the technology that was sold to us as the thing that would make our lives easier is making it multitudes more difficult.

AI detection tools are notoriously flawed at best, ruinous to new authors at worst. They also seem like a particularly inelegant solution to the problem: why would we use AI to catch AI usage?

It’s a tough time to be a reader. Now when I encounter new writing from a name I don’t know, I have to remind myself to be cautious, just in case someone is trying to trick me. The book industry should do absolutely everything necessary to prevent readers from feeling the need to come to any new book from a place of skepticism. Horror critic Emily C. Hughes put it so aptly when she broke down the cancelation of Mia Ballard novel over suspected AI use: “I hate feeling like my choices are to either let myself be hurt repeatedly or to progressively wall myself off into smaller and smaller protective bubbles until I’m utterly alone.” Protect readers at all costs!

If anything, AI detection should fall to book editors, who absolutely didn’t sign up for this new responsibility either.

Our current conundrum is a great reminder of how disastrous it is when corporations rely on technology instead of investing in human expertise. Because at its heart, AI and the havoc it’s causing is a labor issue.

AI paranoia is a symptom of a larger problem: a corporate culture that values quantity over quality.

Editors, for the record, are already overworked. In fact, just about anyone who has a job at a corporate publishing house is overworked. When publishers cut staff, which seems to be a near weekly occurrence these days, the people who remain have to take on all of the work that their former colleagues can no longer do on top of the work they were already doing. Doing the work of two or three people in the same amount of time you always had means that nowadays editors are short on time to actually edit.

Add to this job security that is shaky, with an editor’s worth always depending on their book sales track record from the previous year. It’s not an ideal environment for productivity, let alone for making art.

AI paranoia is a symptom of a larger problem: a corporate culture that values quantity over quality, quick output over the very human labor of poring over a manuscript and making sure that each and every word sings. This is especially true for self-published works that are acquired by major publishers, works that may not have been rigorously edited the first time around but deserve a more thorough reading before thousands of copies are printed.

Editor is the role in the publishing process most equipped to work closely enough with a text to know, if not when AI is being used, then at least when a word or a sentence construction feels off. The more time an editor has to edit a particular book, the more care they can put into it (this is why most of them got into the book business in the first place!). The more care they put into it, the more likely that they can deal with clunky writing of any sort, which, at best, would include any attempts to pass off AI as original writing. Giving editors more time and space to, you know, edit, is not a definitive solution to the AI problem, but it’s a good start.

Sorry if this is stating the obvious, but humans aren’t machines that can churn out work unceasingly without rest, especially when the output is creative. In fact, we’re at our best when our working conditions allow us to spend quality time thinking, contemplating, analyzing, interpreting. Books aren’t widgets, after all (or at least the good ones aren’t). They require the guiding hand of editors who are empowered enough to be able to take all the time necessary to shape them.