March 24, 2026
The publishing world was rocked last week when a Hachette book titled Shy Girl, by Mia Ballard, was canceled after suspicions that it was AI-generated were raised by The New York Times. Book nerds will know that YouTuber Frankie, a book reviewer, was first to make the claim, alleging back in January that the book was probably AI-generated.
In a since-deleted comment on Frankie’s video, Ballard wrote, “Someone in my writing group offered to help [edit the book] and assured me she’d do a thorough job, so I trusted her. In the process, she also changed a lot of the wording and encouraged me to lean more poetic because that’s my background, and I listened … I did not use AI to write this book. What I can say is that the version you’re referencing was edited by someone else, and I only later realized she may have run parts of it through an AI tool during her editing process.”
There’s no way to know for sure if what Ballard says is true or not. But after several exhaustive reviews of Shy Girl by online book reviewers, the claim that it wasn’t deliberately written using AI (versus just being heavily edited using AI) hasn’t seemed to convince many critics. Once the specter of AI is raised at all, trust in the provenance of the work is difficult to regain. Initially, Shy Girl had good reviews and healthy sales in the UK (1,800 print copies) before the book was pulled. Several blurbs from other authors with solid publishing track records also graced the now-deleted publisher page for the book. So this has shaken the confidence of some in the book publishing space about the future of writing as a profession.
And because of this embarrassing incident, publishers are likely to be more circumspect regarding newer authors, a terrible outcome in a field that is already difficult to break into under a major book publisher. Currently, Amazon is flooded with AI-generated books of fiction and non-fiction, mostly self-published. So when even a professional publisher fails to filter out an AI-generated book, some book fans are wondering, are AI books just inevitable?
I think there’s a different, and more important question to be asked. It’s not, “Can AI increasingly write books and essays that people will read and accept as human-made?” The more important question is, “Do humans care whether or not a book or essay they read comes from a human, or from an AI algorithm?” So far, I think the answer is still on the side of humans.
Humans write pulpy, cliché books all the time, so it’s only natural that an AI-generated book could be passed off as human. But it’s when readers find out that the book didn’t come from a human that they seem to feel duped. This may change, or not. But in the meantime, AI isn’t going anywhere, so what can writers do? Ignore AI? Give in and just start generating more piles of AI-made books? Or do, something else…
The Real AI Writing Questions We Should Be Asking
For a moment, let’s take a leap of faith: What if Ballard was telling the truth? What if she did write the book herself and then gave it to a would-be editor who simply allowed an AI model to over-edit the book to the point where it was effectively transformed into a mostly AI-generated book?
I suspect this has already happened in a number of underexposed cases and will continue to happen in the book publishing industry, mostly undetected. For now, let’s separate those who are deliberately using AI to generate books and slapping their names on the cover. What about authors who, in a bid to be more efficient, fast, and tech-forward, are using AI to help them write and edit their books? These are the people who need the most attention right now, and I think I can help. In 2022, I immersed myself in the world of generative AI images.
Then, in 2023, I began investigating how AI writing works. That led to research into AI writing detection systems, patterns and tropes of AI-generated text, and what this might all mean for the future of writing. But it wasn’t until 2024 that I felt that I had enough information and experience with AI tools to definitively say: I don’t want to read AI-generated writing. I don’t care how good it ever gets. Unless it’s an instruction manual for a product, I have zero interest in reading AI-generated text. But in some places, AI-generated writing is becoming unavoidable.
In the same way that I might drink a beverage and enjoy it, but spit it out if I find out it contains NutraSweet, I’m sure I unknowingly read AI-generated text all the time. But I don’t want to. So, in a time when all of popular culture seems to be pushing everyone who uses a screen to use AI, how can we help writers to use AI in a responsible way that allows us to get the efficiencies of AI, but with real, unadulterated, non-processed, farm-fresh human-written text?
I think the guide below can help. If you are a writer who uses, or is thinking about using AI to assist with your work, I encourage you to use the guide below. And if you’re not, send it to a writer you know who might get something out of it.
THE WRITER’S GUIDE TO USING AI RESPONSIBLY
So let’s get to it. You don’t hate AI, but you do want to use it to optimize your writing process. What does that look like in practice?
First, understand that AI wants to write like the median of whatever style of writing you’ve prompted it to write. If the AI chatbot’s writing has flair or attitude, that only arises because you prompted it. And even then, that flair and attitude will become routine and pedestrian as the word count grows. Will AI get to a point where this isn’t true? Quite possibly. But it’s not there yet, and even if it were, letting a machine express your inner thoughts for you is best left to those who don’t have much to say in the first place. If you think you have a unique perspective and voice, use it. But that doesn’t mean you have to hate or completely abandon AI.
1.
Write as loosely (or tightly, if that’s your persona) as you want to. Don’t worry about mistakes, grammar, or spelling, just write. If you’re experienced as a journalist, essayist, or fiction writer, you already know your process, you should continue following it as you get your first draft finished. If you’re not an experienced writer, then always remember that, unless you’re writing free-form poetry, outlines are your friend. And yes, you may be tempted to use AI to help you craft your outline, but I’d advise against it. I’ll offer two exceptions to that rule:
A. If you’re an experienced writer, and you know what you want, and have the skill to maintain your own chain of thought, then no matter what outline an AI chatbot suggests, you’ll probably be ok.
B. If you’re not experienced, but under a tight deadline, an AI-gen outline can be helpful (but like any vice, letting it become habitual can be problematic). After you get your AI-generated outline, you may be tempted to just let AI write the rest of the work for you, but don’t. Only use the outline (which you should really be writing yourself), and then write the actual piece yourself.
2.
Once you’re done with the first draft you’ve written (remember, no AI, just you roughing it out on the page!), give it a read. Does it make sense? Are you telling a coherent story and laying out an easy-to-follow narrative, or are your thoughts all over the place? If it’s the latter, then you probably needed a better outline, assuming you used one at all. On the other hand, if you like the general thrust of what you’ve written, it’s time to go in and firm up any weak points. These can be run-on sentences, weak adjectives, and areas where your logic, arguments, or descriptions simply aren’t as clear as they could be. That’s your second pass.
NEWBIE NOTE: I’d like to remind readers here that I’m primarily speaking to experienced writers. If you’re an inexperienced writer, there are many other suggestions I’d offer for writing in general. In this case, I’ll keep it short and say that if you’re an inexperienced writer, repetition is your friend. And by repetition, I mean that you should write as much as you can. Similarly, you should read the work of other writers frequently. You’ll be amazed by how reading a writer whose work you appreciate can function as literary rocket fuel and spark new enthusiasm and a fresh perspective in your own writing. Now, back to experienced writers.
3.
Assuming you’re generally happy with your second draft, now it’s time to give AI a look at it. And this is where things get tricky. You must remember that an AI chatbot cannot give you a writing style or a literary personality. It is, essentially, a median expression of everything it has been trained on. It has no style, no soul, no real voice. AI can only mimic such things. So if you want your writing to “sound” like you, you must retain agency, ditch any imposter syndrome still lurking in the back of your writerly mind, and assert dominance over the AI chatbot’s edit suggestions.
Got it? Good. With that in mind, you’ll first want to use your AI chatbot to check for simple errors of spelling, grammar, and, when you’re more advanced, potential weak points in your logic that could use additional clarity. Also, it can be useful to use two different AI chatbots at this stage. Why? Because AI isn’t perfect, and what one AI chatbot misses, the other may very well pick up. Take your second draft and enter it into the AI chatbot with the following prompt.
Prompt: Check the following essay for spelling and grammar errors. Do not rewrite, list any corrections.
This is where many people using AI chatbots for editing their writing get tripped up. Instead of using the above prompt, they allow the AI chatbot to rewrite their work while including the corrections. So instead of just correcting your mistakes, the AI chatbot also rewords your essay. This is IMPORTANT. You cannot let this happen. Your rule should be: I will not allow the AI chatbot to rewrite a single sentence, ever. If you allow this, you’ll eventually begin to allow more and more rewrites and, next thing you know, at least half (or more) of your essay has been written by AI.
For a busy and inexperienced person, this may seem like an efficient way to save time while “sort of” retaining an authentic fingerprint as the author of your work. It’s not. If you do this, you’re lying to yourself. Once AI rewrites your paragraphs, your writing begins sliding toward that median of all writing, with just a little (increasingly indistinguishable) you seasoning here and there.
You must remain disciplined and only use the AI chatbot to find your misspellings and grammar problems. In fact, I’ll even suggest taking grammar corrections as “suggestions,” not law, as grammar changes can also transform the nature of your writing. Only change grammar if you must. Dialogue, writing rhythm, and cadence as style, as well as personal flourishes unique to you may all register as “bad grammar” with the AI. You must work to discern what is wholly incorrect grammar, and what is the AI simply disagreeing with your rough human edges that give your writing personality.
Also, notice that I didn’t include punctuation correction suggestions. Although AI chatbots can save you from glaring punctuation mistakes, I’ve found that it’s just as likely to em dash and semicolon you to death as yet another way to median your writing into an AI chatbot voice instead of your own human voice. You’re not aiming for perfection, you’re aiming for coherent, easy-to-read writing that retains your voice and is free of almost all errors.
Some errors are good. Slang AI doesn’t understand. Cutting-edge colloquialisms not yet sucked up from social media into AI models. And sometimes you may want to deliberately make up words, be it a new portmanteau, or just something funky that you think is fun and that your readers will understand. AI doesn’t want to do that for you, it wants you to be median. A recent research study published earlier this month by UC Berkeley, UC San Diego, University of Washington, Zaytuna College, and Google DeepMind found that, “When LLMs revise human writing, they induce large homogenizing changes very unlike how people would have edited the same essay.”
Your writing must represent you, not a perfect vision of you as dictated by a Large Language Model. It must be you who adds the sometimes ugly and unique touches to your work that will make it undeniably yours. If you want to write like elevator music, fine, ignore all of this. But if you want to write like a human with a soul and a point of view, this is all incredibly important to keep in mind when engaging any kind of AI tool.
4.
Once you’ve corrected everything, and shunned any rewriting suggestions, you can now take your essay into AI chatbot number two, preferably one from a different organization with similarly robust models. Use the same prompt: Check the following essay for spelling and grammar. Do not rewrite, list any corrections. And watch with wonder as you (sometimes) find that the second AI chatbot found very obvious errors that need to be corrected that the first AI chatbot missed. Make your corrections, and you’re almost done.
5.
Now, go back and slow read your written work again. This time, you’re not only looking for errors that the AI chatbot missed, you’re also looking to see if any of the corrections changed the meaning or tone of your work. And here I must emphasize again, you aren’t looking for perfection (there are no perfect writers, just writers who have perfected their own voice and style), you’re checking to make sure your writing still has the color, flavor, and energy you meant it to have at the start. If you’ve followed my directions, and didn’t allow any AI rewrites, only technical corrections, then any tweaks, if any, should be minor. This won’t guarantee that you’ll produce a great piece of writing, but it will ensure that the writing is yours. And if you need to improve (as we all do), using this AI-assisted method, you can do so organically over time and with practice.
And there you have it. That is how you can use AI chatbots to become a more productive and effective writer while remaining the true author of your work and without surrendering your voice to a machine.
FINAL NOTES:
1. I used to love Grammarly. But now, I find that it too frequently attempts to hijack the writer’s voice with aggressive suggestions and corrections that change the foundation of how a piece reads. For example, if you’re a college professor, you may think your student’s work reads like AI, or maybe you ran it through an AI checker and found a troubling result. While it may indeed be AI-generated, you should also keep in mind that the student may not have prompted an AI-generated essay, they may have actually written the essay, but foolishly allowed an AI-assisted tool like Grammarly to take control of their edits. In some cases, these edits reach the point where the essay becomes riddled with AI-rewritten passages that will be flagged as AI by an AI checker.
2. I don’t think AI chatbots are necessary for writing in 2026. If you are detail-oriented, get cozy with Strunk & White or Garner’s, or have a good human editor available, you’ll be perfectly fine. I’ve put this guide together to help the growing number of tech-savvy writers who either need to harness AI to speed up their work processes or are AI newcomers who are fumbling their way around AI tools and learning bad habits. This guide can help writers to avoid making mistakes with AI that will rob them of their authorial agency in favor of an algorithm.
3. AI checkers are still not universally reliable. As of now, these tools cannot meet the Daubert standard of scientific evidence that would stand up in a court of law. However, some AI checkers are better than others. I’ve seen a viral social media post going around mentioning that President Abraham Lincoln’s Gettysburg Address came back as “likely AI” when run through an AI checker that I won’t name here. None of them are perfect, but the AI checker that has delivered the best results for me, even when I try to trick it, is the one from Originality.ai. It gave Lincoln the 100% human score he deserved.
4. All of this was delivered in good faith to people who are or want to be writers. For those out there determined to use AI to generate essays and books, I’m sure you’ve read all this for counterintelligence purposes on how to “beat the system” with AI-generated text. In the end, I think you’re devaluing your own experience and value as a human, even if your sole aim is just to make money.
But let me add another piece of information for those who plan to use AI to write for them: None of those “humanizers” that promise to scramble, adjust, and distort your AI-generated text so that it reads as more human when sent to an AI checker actually work consistently. For now, it’s a waste of time and money. Will humanizers get better? Almost certainly. But like I said, even if the AI humanizers become good enough to fool AI checkers, you’re robbing yourself of something important, a meaningful human experience. And if you don’t know the value of that, then there’s nothing I can say to convince you otherwise.
New essays on AI, film, & entertainment innovation arrive by email. Subscribe to stay in the loop. All editorial text is written by humans.
Cover image: “Kolchak: The Night Stalker” via (NBCUniversal)
