The end of books as we know them

This is not an argument for artificial intelligence.

This is not an argument against artificial intelligence.

At least here, I am not interested in:

• Passing moral judgment
• Demonstrating my belonging to one side or the other
• Engaging in doomerism

This aims to be a fairly clear-headed and dispassionate post about what awaits us in the near, possible — if not probable — future.

Premise Number One:

Imagine being an audience member at a magic show. The magician pulls a rabbit out of a hat. Applause. Once the rabbit is out, it hops freely across the stage — and maybe even onto the guests.

Now, is it inconvenient for the rabbit to jump onto the audience's laps? Maybe. But it’s not going back into the hat. It cannot go back.

The trick has been done.

In this analogy, the rabbit is LLM (Large Language Model) technology, and the hat is the void. Now that we have discovered it, we cannot forget it.

All legislation regarding AI (which, by the way, is necessary) must keep in mind that a trained model can now run on anyone’s home computer. Not even a surveillance state could prevent this.

Premise Number Two:

Let’s assume technology continues to improve.

It doesn't have to improve rapidly — it could progress slowly, even with occasional setbacks. It doesn’t matter.

Already today, it is possible — both theoretically and practically — to write a book using AI. In fact, some people are already doing so³.

On her Substack Worth Writing For, Olivia talks about how her roommate is having GPT write a book. The process is still slow and clunky (requiring multiple prompts), but already dozens of times faster than a normal writer.

What’s interesting about Olivia’s post is that, as far as we know, the roommate doesn’t even intend to publish it — she’s doing it just for her own amusement.

Olivia’s roomate might be well intentioned, but we know this is not always the case.

Publishers already make use of AI.

Spines is aiming to release 8,000 books a year using AI, and their outspoken, brash and kinda nonsensical approach might be just the visible peak of an iceberg. In Italy, the publisher Tlon published hypnocracy (Ipnocrazia), the first essay from Jianwei Xun, just to later reveal Jianwei Xun doesn’t exist. The whole thing was AI written.

It might be tempting to condemn these cases. But personally I’d be far more worried of the unseen AI use.

Publishing houses are businesses concerned with profit.

Right now, profit means pushing as many releases as possible onto the market at the lowest possible cost, with minimal risk. In this context, not using AI to streamline the process would be a very stupid decision.

This is obvious — but we must remember that OpenAI, Claude, Sudowrite, etc., are not industry-exclusive tools, hidden away in a publisher’s office.

They are available especially to end users. This is crucial. With OpenAI being so outspoken and out, publishing houses cannot keep the technology in the closet.

In five years, we'll have "on-demand" novels.

Right now prompting a whole book is cumbersome. But if you accept my premises, in five years it might be as simple as writing a sentence and pushing a button.

Imagine your book - a command away. On demand, meaning on request.

Any user could ask an AI:

Write a novel inspired by The Lord of the Rings, where the hero is chosen by destiny but doesn’t know it, and make sure there are dragons — and that they’re good.

Sounds familiar? It should, it's the premise of countless modern fantasy books.

Or:

Write The Winds of Winter, the book George R.R. Martin never finished.

You press Enter, and after an hour, five hours, or a day, the model hands you a complete book — ready to print, turn into an ebook, read, illustrate, or narrate in audio form.

Sure, maybe it would be terribly written.

Two years ago, when a writer friend introduced me to Sudowrite, I wasn’t worried. My friend was maybe more far-sighted than I was, for he was already sweating. I tried the tool, saw its limitations, and concluded it was just a tool.

Today, both Sudowriter’s own Muse model and GPT-4o can produce prose at human level (provided the prompts and editing are right). You'still need major oversight in order to smoothen out the most glaring example of AI slop, but it’s incredibly serviceable.

Maybe it’s not saying much, but both tools produce better prose than I did two years ago. Sure, I’m a non native speaker, and I was barely on the English market ever, yada yada. But the argument stands.

If it keeps improving in any way, then it’s reasonable to think that "assisted" writing will be able to replicate any writer.

Quality isn’t the real problem.

In The Colors of Her Coat, Scott wonders if being able to generate 10,000 "Ghibli-like" images on your home computer cheapens the experience of watching My Neighbor Totoro or any masterpiece from Studio Ghibli.

If you insist that anything too common, anything come by too cheaply, must be boring, then all the wonders of the Singularity cannot save you. [...] But if you can bring yourself to really pay attention, to see old things for the first time, then you can combine the limitless variety of modernity with the awe of a peasant seeing an ultramarine mural — or the delight of a 2025er Ghiblifying photos.

I’ll let you read his full article (it’s fascinating).To me, it suggests that the banalization and degradation of the artistic process are less about technology and more about — sigh — consumption.

An on-demand book would be incredible. No need for distribution, editing, bookstores; it would be tailored exactly to the user’s desires.

Which inevitably brings us to the subject of:

The Death of the Book

In a world where any story is just one prompt away, the book as we know it loses purpose.

Already today, readers can get the same kind of satisfaction from dozens of other media (video games, movies, series, podcasts, anime, comics, etc). In the near future, they could fulfill their reading cravings instantly by generating a custom book.

No more waiting years for the next book in a series. No more fanfiction. Bow to the infinite story engine.

We already know that the major bottleneck for reading isn't the lack of material (bookstores are overflowing) — it’s the lack of time.

I'm sorry to tooth my own horn here, but we all compete in the attention market.

A reader busy with “The Chronicles of Simple Fantasy” won’t have time to pick up “The Unknown but Beautiful Book” because their limited time is already occupied.

Being able to generate your own perfect book (themes, plot, setting, style, characters) would introduce ruthless competition — not over money, but over time and attention.

The mere existence of affordable AI-generated stories could shrink or saturate a portion of the market even without ever getting in bookstores.

But the death of the book mostly affects writers.

How long does a professional writer need to produce a first draft? Six months? Three?

Too long, when a machine can do it in days.

Most writers aren't even professionals; they can’t or don’t manage to sustain that pace because they don't make a living.

Spending a year (counting revisions, editing, marketing) with zero guarantee of success is already hard today. Most writers don’t even know if their novel will see the light of day.

Writing could still exist as a hobby — like painting portraits, building models, or playing soccer without expecting fame or reward.

It would be an art without an audience.

Photography "killed" portrait painters. Trains and cars "killed" horses.

Similarly, books wouldn’t disappear, but could shrink into a tiny niche, like avant-garde literature.

How much avant-garde literature gets read? How much indie literature?

If your answer was “not much,” you’ve won a bonus question: can we expect that niche to grow in the future?

A False Solution: Certifying "Made by Humans"

Some associations propose “certifying” books as handmade by human authors.

No tool today can certify AI-absence without false positives and negatives. The worst tools are standard statistical algorithms; the best (ironically) are AIs themselves.

In short, these initiatives are mostly meaningless.

In any case, we could at best end up with an AI vs. anti-AI arms race. At worst, we’ll have witch hunts — as already happens to illustrators, who are constantly asked to prove their work is theirs.

More than a year ago I was subjected to one of those false positive cases, where the algorithm here on Vocal suspected me of pushing AI-generated stories on the platform.

Without a real technical guarantee, any certification is just empty moralizing.

The issue has always been the same.Technology grows exponentially; society’s ability to adapt grows linearly.

What if the Book Dies?

I believe humanity has always had the impulse to create — or more specifically, to tell stories — ever since we painted on cave walls.

Even if the book dies, creativity will find new forms. That doesn’t worry me.

What worries me, as a writer, is working in a storytelling format that could become obsolete within five years.It worries me to enter the AI debate clinging to outdated ideas.

There’s enough material for another essay here.But in any case, if you’re in the market, you have to ask yourself what it means.

Everyone has dreamed at least once of "making it big" through writing; but maybe that dream was never realistic.

As for the "book problem," the real question is: how will storytelling evolve? What new ways might emerge to tell stories?

I honestly don’t have an answer. My prophetic ability stops here (if it ever existed).

On-demand books, as I described, have a major flaw: they would lack a social aspect. Readers could generate their own stories — but without sharing, it would be hard to talk about them.

Moreover, there’s the typical "black box" issue of AI, potential human skill loss, and lots of philosophical and practical implications — which, again, are no longer theoretical. The rabbit is out of the hat.

On a more positive note, AI could empower collaborative storytelling — for example, in role-playing games, where AI could provide settings for two human explorers.Solo RPGs, which rely heavily on journaling techniques, could also benefit from these "idea engines" or "mechanical narrators". Maybe we’ll see a rebirth of gamebooks.

But most of all, AI might open the door to forms of storytelling we can’t even imagine yet.

Instead of trying to stop the current, it’s important to start an open, honest debate.One that includes both writers and the public.

What if the Book Doesn’t Die?

I promised not to engage in doomerism.

Change isn’t guaranteed.The market is very good at co-opting destabilizing technologies.

In five years, we might have the technical ability to generate novels from a single sentence... and still keep going to bookstores.

People read little now, and may continue to read little.Some jobs linked to the book production cycle (editors, evaluators, ghostwriters, proofreaders) could take a big hit.

But readers might still rely on publishers’ books, trusting (more or less) their internal processes — or simply letting marketing guide their Christmas shopping.

As Giuseppe Tomasi di Lampedusa wrote in The Leopard:

"If we want things to stay as they are, everything will have to change."

In short, everything could end in massive stagnation.Which might be an even worse outcome than the death of the book.

The Great Unspoken

Publishing is in a crisis.

Any discussion about the future must start from this simple fact:Massive amounts of printed books are produced only to be pulped. Admittedly, I know this very well from my encounters with the Italian market. It might be a smaller, particularly terminal branch of the economy, true enough. But I believe this section to be true for the larger English market.

Printing millions of books to recycle them later should make you laugh. Imagine a printing press throwing rotaries straight into the recycling bin.

The reality is worse.Before being unsold, books must be transported, distributed nationwide, stored, displayed, returned — and then destroyed.

More than laughable, it should be horrifying.

The publishing industry’s response so far has been to print even more books in a mad race to compete.

ChatGPT, like a good non-thinking algorithm, will serve both the solitary user and the series editor eager to accelerate a publishing machine already on a collision course.

In conclusion: if the book dies, it won't be because of the new arrival.

They’ll say it had been sick for a long time.And maybe — just maybe — we’ll realize it was poisoned.

original post appeared in Italian over Il sindacato del Demone.