It’s okay to not use AI. It’s equally okay to use AI, as long as you know where it actually helps and where handing over too much of the writing hurts the work.
But what is not okay is dreaming that AI would do your job for you or dismissing it altogether. I see AI like the internet. It’s going to be here whether we like it or not. So it’s better to start learning how to use AI right. By the way, the internet did open a lot of doors for us, both good and bad ones. We just had to learn to spot which is which.
When I sit down to write something, AI benefits me the most during the first 20% and the last 10% of the task. In all of those tasks, AI gave me a head start; it helped me round off the rough edges toward the end; but the in-between 70% of the work had to be mine to achieve the quality I wished for.
So I don’t see LLMs as a panacea, the universal cure for all my technical and nontechnical hiccups. I’ve come to see it as a tool, something I should know when to use and when to stow away safely in my metaphorical toolbox.
AI-assisted vs AI-generated work
There is an important difference between AI-assisted writing and AI-generated writing.
- AI-assisted work is where you befriend AI to help you catch major flaws in your storytelling, deviations in structure, missing core concepts, obvious inconsistencies, or sections that simply aren’t pulling their weight. You still make the creative decisions, and AI only inspects them at the macro-level. Essentially, you write the first draft, AI helps you find major gaps and sometimes also suggests how to fix them, and you evaluate these suggestions and arrive at the end result through your own decision-making process.
- AI-generated work hands the actual writing to the machine. Fully AI-generated writing works only for purely or largely programmatic content like product descriptions, templated SEO pages, summaries, FAQs, and similar business material where readers are primarily concerned with the information being conveyed rather than the writer’s personality or opinion. ANY piece of writing whose goal is to evoke a particular emotion (curiosity, tension, empathy, etc.) in readers should not be fully AI-generated.
Sophisticated AI tools are reshaping industries, and publishing is no exception. From generating plot outlines and character profiles to drafting chapters and assisting with editing, AI tools are changing how books are conceived, written, edited, and produced.
When the integration of AI isn’t just about speed or cost-saving but about exploring new possibilities in storytelling, it can be beneficial.
Where it doesn’t work is when AI becomes your substitute for writing. This is where your typical day of “writing” turns into a chunk of well-intentioned but increasingly desperate prompts to ChatGPT or its comrades:
“Make this character sound angrier.”
“No, not that angry.”
“Make it subtle.”
And finally that prompt of ultimate desperation: “Make it more human!”
At that point, you’re no longer using AI to help you write. You’re trying to direct AI into writing the book you have in your head, which it can’t and won’t.
On the contrary, AI-assisted books can actually work. And for that to happen, authors and publishers need to make sure AI remains a tool in the process rather than the doer.
Making AI work in fiction
For fiction writing, I have learned from experience that it’s okay to use AI to catch major oversights in character development, storytelling, pacing, or the sequence of events. AI can do that; it can notice when an important revelation has little setup or when the middle of your story introduces a character with no context. What it is considerably less dependable at spotting are the small but important shifts that make characters feel real. It may not explain why the angry inn owner who has distrusted the protagonist for the past 150 pages suddenly decides to help them find the murderer, or perhaps why Ms. Clairmont suddenly forgives her toxic boyfriend after the previous chapters established her wisdom and maturity. Technically, both events can be made to fit the plot. Emotionally, they may make no sense whatsoever. These are problems of character interiority and motivation, which develop from details scattered across hundreds of pages. AI can flag some of them but not all, especially when evaluating an entire manuscript in one go.
What AI also can’t do in fiction copy editing, as of the time of writing this article, is catch the frequent repetition of the central conflict. As in any weak, hastily put-together Netflix series that repeats the central conflict almost every other episode without moving in the direction of resolution, AI thinks it’s okay to bluntly restate the plotline/central conflict often because that’s enough to justify everything!
Reiterating the central conflict can make a scene appear relevant to the larger story, which is why a writer may return to this writing technique now and then, but AI certainly will overdo it. These are specific areas where a human copyeditor can do a far better job than any AI model on the market.
A good copyeditor simultaneously tracks meaning, continuity, characterization, sentence-level choices, repetition, tone, and the effect those choices have on the reader. You have a much better chance of arriving at a genuinely polished manuscript by hiring a human copyeditor and using AI alongside the process than by asking AI to copyedit the entire book for you.
Making AI work in nonfiction
For nonfiction, AI can be particularly useful in qualifying whether the chapters are pulling their weight (although not so finely for books over 30,000 words because AI tools are severely limited by their inability to retain context across long-form content), whether a concept lacks supporting material, or whether your book broadly follows the general structure of warm-up → problem → solution → supporting material.
Long-form books remain difficult for AI because good developmental editing requires not just remembering what has happened in the first 10,000 words but also having a feel for the pacing and evaluating if the book actively builds upon its arguments. This should be done in such a way that the reader is neither overwhelmed nor bored by the pace, and that means change is necessary. Tasteful variation is what keeps the reader hooked to a book, not mechanical and monotonous “consistency.”
AI is also not a substitute for proper fact-checking, which surprised me initially because fact-checking sounds like exactly the sort of job AI should be good at. There is a major difference between spotting an obvious inconsistency (such as saying the sun sets in the north) and knowing that one research paper that says intermittent fasting is good does not advocate for the conclusion becoming a fact.
So far, I have seen AI confidently accept:
- a plausible but incorrect statistic;
- a study that does not actually support the conclusion attributed to it;
- a confused correlation-and-causation argument;
- a medical, scientific, historical, or technical claim that sounds plausible but only in a highly limited environment and when used with a lot of disclaimers!
AI can also struggle with another important nonfiction editing problem: whether the book has genuinely proved what the author thinks it has proved.
An argument can be coherent, well structured, and beautifully written while still relying on weak evidence, unsupported assumptions, or examples that don’t justify the broader conclusion. That’s exactly when a human copy editor or developmental editor can offer you a “sensibility check” for your manuscript.
Making AI work in business content
For business content, AI is remarkably capable of producing programmatic content with near 100% accuracy and almost no bias. The best places to deploy AI are structured, repeatable, information-heavy content, like product descriptions, internal summaries, FAQs, catalog copy, or standardized emails.
But for creative content or anything that you want your clients to engage with, it’s better to use AI-assisted content, in other words, human-led content, rather than purely AI-generated content. If you’re writing a founder story, thought-leadership article, sales page, customer case study, brand campaign, or anything intended to persuade clients or make them feel something about your company, purely AI-generated copy often falls short in doing the task.
The verdict…
For nonfiction, use AI to challenge the argument. Ask it what you have missed. Let it find repetition. Make it suggest alternative structures. Ask it to attack your assumptions. Then write the thing you actually want to say.
For fiction, use AI to identify gaps in your character arcs and plot development. Then proceed to fix them on your own, or you can even ask AI to assist you by suggesting options.
For business content, use AI only for programmatic, information-heavy content. For anything that needs to evoke an emotion (curiosity, interest) or an action (purchase, subscribe, contact), go with human-led content.
Using AI in your writing doesn’t mean handing over your voice. If your AI-assisted manuscript needs a human editor to catch weak arguments, repetitive phrasing, flat passages, or places where the writing simply doesn’t sound like you, get a free sample edit from us. See what professional human editing can do for your book.
Leave a Reply
You must be logged in to post a comment.