1. Introduction
AI in Literature… Is it worth it? (no).
I’ve done a deep dive and investigated the ethics, quality and legalities of AI in literature to be more educated when I tell people precisely why I would never use AI for this purpose. And in my investigation, I used AI systems with a critical lens, so that you don’t have to.
See, I want to write, it’s the writing and crafting the stories that is my hobby. I don’t want to outsource my writing to a third party. Why would I give the fun part away?
In the spirit of transparency, my position is this:
- AI is a fundamental and overall net harm to the literature ecosystem.
- The few upsides are vastly outweighed by the downsides.
And I am going to use this article to prove my position.
What am I not going to discuss in this article?
Anything criticism that can be levelled at AI as a whole. The things such as theft of intellectual property, environmentalism, third party thinking and of course, the fact that it’s almost exclusively being used by corporations as a way of reducing headcount. There are so many negatives, but… if I may, I’m going to focus on discussing just its impact on the literature ecosystem. This article is intended to discuss the problems that are more often than not unique to generative AI in literature as opposed to other domains.
Also, just as a point of housekeeping – when I refer to AI, I’m actually just referring to Large Language Models, or LLMs. ChatGPT, Gemini, Claude, etc. As I’m sure we’re all aware right now, these aren’t thinking machines, despite what these companies will try and get us to believe. They simply are good at predicting sequences of words to varying degrees of complexity.
2. Who Uses AI for Writing? The Target Audience
First of all, let’s get our bearings.
- Who is this for?
- Who is the target market for technology that allows you to write stories without writing?
In my research, I can point to two different personas, representing two different types of people that AI writing appeals to. If you’ve started reading this article with a positive view on AI in literature, then you are likely one of these people:
The one who wants to be a writer, but has little interest in writing itself.
This is the person who wants to reap the rewards of being an acclaimed author. They want the fame, the accolades, the adoring fans, the Booktokkers swooning over them. It’s just a shame that in order to get there, they have to actually be a writer.
They want the glamour, they want the paycheque…and in that, this persona is doomed. Very few authors make significant money from their writing these days, and a large number of popular published authors still maintain their day jobs. There’s just not enough money out there for authors. From the outside looking in, you might be able to see the glamour, the book signings, the thousands making fanart, the convention appearances and think that this means they’re rich. It’s not the case, deceptively few authors make this type of money.
The one who does not have the will, or energy to learn the craft.
This is the one that is most representative of this day and age. The person who wants it right now: the instant gratification user. These users may love the craft of writing, but they’re refusing to engage with it out of a sense that they want it now now now, screaming, “who wants to take 10 years learning how to write properly? I want to write and release a book today!”
Now, sneering characterisations aside, there is a case to be made for it being used to make writing more accessible to those with disabilities…but I only ever hear one of those guys talking about that as a gotcha point, so until someone with a disability enters the chat, I have to take that particular point with a pinch of salt.
“So what?” You might be thinking. Why does it matter that I can point to these two types of people to whom AI writing appeals? Well… here’s the crux of it:
Those two personas are dismantling the literature we all know and love.
And I’m going to talk about specifically how.
I want you to keep in mind these two personas as we go through this article, because some points will apply to one, the other, or both.
Let’s get into it.
3. Writing An AI Generated Novel
As a bit of a disclaimer here, this will be dripping in hyperbole, but hyperbole is exaggeration, not lies. There is still a strong basis in truth, here. These are my experiences based on my research, and also my predictions and conclusions, not made-up dystopian fantasies. I’m attempting to give you a taste of what may be around the corner, but I also acknowledge that in many cases, we’re not quite there yet.
3.1. How AI Homogenizes Literature and Kills Creativity
Large Language Models are, by their fundamental nature, word prediction engines, not word crafting engines. Their job is to look at a preceding series of words and predict the most likely next word. They are simply auto-correct. We’re asking auto-correct to write novels for us. Which is madness. Granted, I acknowledge that this is an oversimplification of what is quite a complex ecosystem.
It’s this ‘most likely’ phrase that kills literature. Because when things become more likely than others, the end result can only ever be homogenisation.
What we will see is a flattening out of all prose, where instead of seeing creativity expressed through the authors voice, we will see those ‘rough edges’ buffed out until the writing is unblemished and smooth.
Now, the way I’ve phrased that almost makes it sound like a good thing, right?
Well, it’s not.
True creativity comes not from the ‘average’ or the ‘homogenous’, but those that stand consciously defiant of such averaging. The most visually compelling sensory descriptions are the ones that stand far away from the expected description for that phenomena, the ones that force us to make connections between different sensory experiences. The cadence or flow of a sentence that stands at odds to the rest delivers a specific experience to the reader. All of these things combined are what makes literature great. They challenge our minds.
What we are seeing here is a reduction in challenge. We will no longer be challenged by an author’s voice, no longer challenged by things that are non-standard. Because everything will be standardised, everything will be the same, it will all be averaged.
And literature loses its heart.
3.2. The Problem with Hyper-Personalized AI Stories
One of the ‘pros’ I hear many speaking of with AI writing is that it allows for us to have stories specifically tailored to our interests. To have a story that feels like it was written for me, and me alone. And I can see the appeal of it, for sure.
However, allow me to alter the lens of your perspective. Just tweak its focus a little.
One of the best things about writing is the community around it. Sharing our favourite stories with one another, finding friends and bonding over what we’ve all read…it’s the very thing that captures us as people. Literature allows us to connect with others, to share experiences, to develop empathy for one another. Reading outside our interests and then sharing that with a group is a core moment for the development of you as a person.
Now, what do you think is the natural consequence of having books that are written for you and only you?
For one, you’ll never read outside of your interests again. Why would you need to? There’s infinite books that can satisfy your exact needs, and if not, you can write more, right? So you will lose empathy, you will lose the ability to have your perspective shifted or your mindset challenged. Literature isn’t just entertainment, it’s a way to grow and better ourselves. But now, you don’t need any of that nonsense, you can just read endless copies of the same books with you as the self insert character where you win against all odds again and again and again.
The second point here is that you’ll lose the community of readers. You’ll never share your experiences with another person ever again. If everyone’s literature becomes personal, then there is never a cause to share books, to share the laughs, delight together in the plot twists, or recommend books to friends. I don’t want to read your self insert personalised story, I’ll have my own self insert personalised stories.
There will never again be any genre defining greats, because we’ll all have our own genres, and we’ll never speak with another reader about them.
This hyper-individualisation of our media is pushing us all further away from one another, and making us lonelier as a species. We should be looking for opportunities to connect, opportunities to bring us closer to one another, not to isolate ourselves further from communities.
3.3. AI Copyright Laws: Can You Copyright an AI Book?
Is the prompter the author? Does the prompter own the book? Does the prompter own the copyright?
Here’s the thing, legally speaking, a book written by an LLM is considered Public Domain work… At least as far as courts in the US have ruled so far.
Other jurisdictions have similar laws, and some have opposing laws, but as of the time of recording, not enough jurisdictions have ruled concretely in one direction to give an indication of global consensus. But, at least in the US, the US Copyright Office and the Thaler v. Perlmutter ruling states that one cannot copyright AI produced works at all. They ruled that “human authorship is a bedrock requirement of copyright.”
If I was a touch less ethical, I could download the ebook for an AI written book, put my name on it, and release it as a competitor to yours. And there is nothing you can do about it.
There are of course some caveats to this: for example, if you change, or alter the work, then ONLY the pieces you change are copyrightable. And then you’d need to be able to prove, without a shadow of a doubt exactly which bits are yours and which bits aren’t.
And of course, with such exhaustive line-by-line evidence, I could just rewrite your bits and then release it again.
The masterpiece you create with AI, is now my masterpiece, and there’s not a damn thing you can do about it.
The only way to protect yourself fully, is to rewrite it all yourself…but as we’ve already established, the Venn Diagram of people who are willing to write the whole book, and people who write with AI are two separate circles.
As time goes on, there will be more legal challenges of this nature and I’m proud to say that so far, the courts seem to be siding with those who have had their work stolen.
4. Can You Use AI To Support Your Writing?
4.1. Using AI Tools for Developmental Editing
What is developmental editing? Broadly speaking, it’s having an expert in their field look over your manuscript to check for plotholes, character arcs, narrative arc, pacing, and broad-brush things of that nature. This is distinct from line editing or proofreading in that it doesn’t usually care about the spelling/grammar, or line-level structural concerns. It’s looking at the story as a whole, rather than granular edits.
LLMs, being large language models, are actually pretty good at giving you developmental editing feedback. It’s able to read and draw conclusions from the text, using only the text as its source context…except for a flaw I’ve found that invalidates the entire experience, rendering it pointless.
Developmental Editing with the aid of generative AI is like working with a toddler who happens to be an expert in developmental editing.
“What the hell does that mean?” I hear you ask.
One of the reasons developmental editors are so good at what they do, is that they have a firm set of beliefs and standards, and what they’re doing, in essence, is determining whether or not your manuscript adheres to, or veers from those standards. Whether those beliefs and standards are codified somewhere or not, every developmental editor knows what they like and what feels off to them, and they’re able to effectively communicate why that is.
Well, LLMs know the rules of developmental editing…but it can very easily be convinced to abandon them. Imagine you get feedback from your AI engine that the pacing of Part 2 of your novel is far slower and is likely to be a fall-off point for readers.
You can just say “it’s supposed to be like that, so evaluate it with that in mind.” and it just accepts what you say, and tells you that you’ve made a very good decision here.
But that’s not what developmental editing is for.
You shouldn’t be able to convince them to change the rules for you, that’s in direct opposition to this entire purpose of developmental editing.
If you’re brave enough and you have a manuscript you don’t mind getting scraped and stored (or you’re paying for it to remain private) then try it. It can be convinced to change any criticism it levels at the manuscript. It may say that the protagonist doesn’t have enough agency between Chapters 1 and 8. You can counter with “that’s the point.” and it will say “Good point, let me re-evaluate based on that feedback.”
You can convince it that it’s perfect, when it’s very obviously not.
A human developmental editor would push back. And that’s the point.
4.2. Using AI Tools AS A Critique Partner
Having a critique partner, or reader is such an important part of the writing journey. It’s a way to keep us in check, a way to keep us from travelling away with the fairies on our writing journey. A good critique partner knows when to gas us up and when to come down hard on what we’re writing… because they have a sense of us, and who we are as authors. It relies on a connection between author, manuscript and reader.
LLMs have no such insights, they can see the manuscript alone.
And so, if you ask for critique it will usually be overly positive unless you specifically tell it not to be. Even if it criticises part of the writing, it will leave you feeling as though you’re doing a great job.
But that’s not what a critique partner is supposed to do.
If you specifically instruct it to be a really harsh critique partner, this also provides a somewhat false lens by which to view the work, as they’re then likely to prioritise giving negative feedback, rather than strong, balanced feedback.
I submitted a short story to a publication, and the editor quite kindly gave me some feedback ultimately disqualifying it from the publication. This feedback is essential to teaching me how to develop my craft.
As an experiment, I ran this same short story through an AI engine and asked it to act as a critique partner.
Not only was the feedback incredibly generous, it failed to pick up on some of the key points that the human editor did. There were a couple of fairly major flaws with the construction of a particular sequence, but these issues were not addressed at all by the AI version.
So, what are they for?
Overall, these support tools make you feel like you’re doing the right thing, but more often than not, the responses are designed to assuage you, the prompter, not improve the output of the writing.
5. So What Is It Good For?
I have waxed lyrical—as it were—about what exactly the drawbacks of fully writing your novel with an LLM, and I’ve also shown how it’s not great as a support tool either.
So, what is it good for?
Well…it allows you to cosplay as an author. It allows you to feel as though you’re contributing to the betterment of the literature ecosystem.
But, just like right-wingers who shout about immigrants, just because you feel like you’re doing the right thing, doesn’t mean you are.
These tools, from what I can see, are degrading the experience for everyone, whether or not you engage with them yourself.
Outside of the literature market, AI is good at summarising data and assisting with research…but there are enough cases of hallucinations or plainly inaccurate answers to give me pause even in this domain.
6. What Is the Future?
I’m not naïve.
I know that the genie cannot be put back into the bottle.
AI is here to stay, unfortunately.
So what do I think the literature market will look like in years to come…
We will see lines drawn in the sand with regards to what can constitute AI-generated work, vs what doesn’t. And this legal distinction, I believe is likely to spill over into the marketplace.
As a result, I see many of the storefronts segregating AI written books. I unfortunately foresee a rise in literature snobbery from those who do write without those tools, as we’ve seen with any large market change when a new technology is introduced.
I see real literature becoming far more esoteric and challenging to read, as a direct response to the overly smooth and unblemished AI writing… if only as an exercise to prove that the author isn’t using AI.
I see stricter laws governing what these companies can scan and learn from…but that these laws will be introduced far too late to have any real effect on those already scanned. I also foresee corporate workarounds to these laws, as we have seen time and time again.
Spotify are (allegedly) already starting to produce music themselves to get out of paying music publishers… and I fear this will happen in the literature industry too. Companies like Amazon, perhaps Barnes & Noble, will likely just opt to write all their novels in-house so they can avoid paying for books from publishers. Publishers themselves will seek to jump on the bandwagon in the name of cost savings…
We will see literature and human authorship reduced to a hobby alone, without a means to make money from it.
Perhaps I’m being overly cynical… but nothing I’ve seen about this environment, this society, has convinced me to be positive about it.
ChatGPT would tell me it’s looking good for the future though…