The Future of AI in Publishing: Automation, Provenance & the Fight for Creative Value
- David Salariya
- 21 hours ago
- 8 min read
Updated: 3 hours ago
There was a time - not so very long ago - when publishing houses were run by people who had ink on their fingers.
They were not always charming. Some were tyrants, some were really horrible. Some were eccentrics. One went to prison. Some were visionaries who mistook obstinacy for genius and were occasionally correct. But they believed, almost irrationally, that books mattered.
I think of the men and women who argued over typefaces, and those who would defend a difficult manuscript because it would “last”. Those who lost money on poetry because civilisation required it. Those who knew their editors by name and their authors by temperament and those who could recite sales figures - but also passages.
They were imperfect custodians, but they understood something fundamental: publishing was not simply a distribution pipeline. It was selection. It was taste. It was judgment exercised under risk.
A publisher was, at heart, a cultural gambler.
Once, a publisher was someone who took risks on books they believed would last. Success wasn’t always certain. Judgment mattered.
Now, many publishing leaders come from sales, finance or rights. They understand scale, licensing and return on investment. Books sit inside larger portfolios of intellectual property. It’s efficient - but it shifts attention away from the messy, human side of making them.
But something subtle has shifted.
Books have become “content units”.Authors have become “talent supply”.Backlists have become “assets under management”.
It is not villainy. It is logic.
And into this logic steps Artificial Intelligence - punctual, obedient, scalable.
A tool that writes.
A tool that summarises.
A tool that illustrates
A tool that does not demand royalties, does not insist upon credit, and certainly never sulks about editorial notes on a colour looking 'cheap'!
To a pipeline-driven system, this looks less like disruption and more like efficiency.
The question is not whether AI will enter publishing. It already has. The question is whether publishing will remember what it once was while it does.
Because if the past century of books were built by those who argued over commas and championed diverse voices, the next may be built by those who optimise output.
And optimisation, as history teaches us, is not always the same thing as culture.
Let us proceed.
Publishing is not one industry.It is trade fiction, academic monographs, children’s non-fiction, picture books, STM journals, educational coursework, licensing, rights exploitation, backlist harvesting, and brand franchises.
Which means AI will not affect it evenly.
Some areas will accelerate. Some will hollow out. Some will harden and
become more premium.
What follows is not panic. It’s pattern recognition.
AI Is Not Replacing Publishing. It’s Rewiring It.
AI does not eliminate creativity.It automates repeatability.
That distinction matters.
In publishing terms, AI is already being used for:
Metadata generation
Blurb drafting
Rights summaries
Cover concept testing
Educational worksheet drafting
Market analysis
Translation assistance
Audio narration
This is not speculative. It’s operational.
Voices
Daron Acemoglu, MIT economist and Nobel laureate, warns that if AI is used primarily for automation rather than augmentation, it risks “creating a permanent underclass” through skill redundancy.
Anton Korinek, University of Virginia, notes that AI’s rapid advancement could cause labour disruption “later this year or later this decade”—uncertain timing, but real structural risk.
Dario Amodei, CEO of Anthropic, has suggested AI could render many forms of cognitive labour economically obsolete within years if development continues at current pace.
These are not fringe voices. They are central figures in economic and technological debate.
The question for publishing is not if AI will be used.It is how it will be used—and by whom.
Publishing Leadership Has Already Shifted
Modern large publishers are predominantly led by:
Sales executives
Rights directors
Finance strategists
Data and operations managers
Books are increasingly treated as:
Assets
IP pipelines
Monetisable catalogues
Brand extensions
This is not inherently wrong. It is structural.
In such an environment, AI appears attractive because it:
Reduces production costs
Accelerates throughput
Enables rapid A/B marketing tests
Scales educational and genre content
Extracts value from backlists
When Bloomsbury licensed educational content to AI firms for training data (reported in 2024), it signalled something important: publishers are beginning to treat their catalogues as machine-learning fuel.
That decision was not ideological. It was commercial.
And it may mark a turning point.
I do wonder about the internal culture of large publishing groups that have grown primarily through acquisition rather than invention.
When growth comes from buying existing lists rather than building new ones, the muscle of creative risk can quietly weaken. Staff inherit formats. They maintain brands. They refresh covers. They update fonts. They optimise metadata. These are legitimate tasks - but they are different from inventing something from nothing.
If the daily work becomes stewardship rather than creation, publishing can begin to feel procedural. Not a vocation, but a role. “Just a job,” as one executive once put it - albeit one she loved.
That distinction matters.
Because when a workforce is trained to manage pipelines rather than originate ideas, a technology that promises greater efficiency within the pipeline feels natural, even welcome. AI becomes an extension of process, not a philosophical challenge.
The deeper question is not whether individuals are creative. It is whether the system they work within rewards creation - or merely maintenance.
When relevance is not actively questioned, automation rarely is either.
The Rise of “Slop Publishing”
Amazon has already seen waves of AI-generated “slop”:
Rapidly produced biographies of trending figures
AI-written travel guides
Low-quality self-help books
Keyword-stuffed non-fiction
What AI-Generated Books Reveal about Publishing
These works are not designed for literary longevity.T
hey are designed for algorithmic visibility.
This is the first visible fault line in AI publishing:
High-volume synthetic publishing vs. curated human-led publishing.
The slop exists because:
Barriers to entry are near zero
Amazon’s ecosystem rewards speed and keyword targeting
Readers cannot always distinguish provenance
But markets tend to stratify. When noise increases, trust becomes scarce.
Scarcity increases value.
Which Areas of Publishing Are Most Vulnerable?
High Risk
Generic genre fiction
Marketing copywriting
Educational worksheets
SEO-driven non-fiction
Background research summaries
Packaged trend biographies
These rely on pattern recognition and recombination - AI’s strength.
Moderate Risk
Mid-list trade fiction
Illustrated non-fiction without strong author branding
Reference books
Formulaic children’s series
Here AI may assist heavily, compressing teams and hollowing out creative ladders.
Lower Risk (for now)
Academic monographs requiring peer review
Highly specialist STM publishing
Deeply researched narrative non-fiction
Distinctive illustrated works
Books anchored in strong author identity
These rely on judgment, authority, provenance, and long-form coherence.
The Hollowing of the Middle
Early signs across creative industries show:
Fewer assistants
Fewer junior editors
Fewer mid-tier designers
Heavier workloads on senior roles
AI replacing entry-level drafting
This mirrors previous technological waves.
But publishing has a unique vulnerability:its apprenticeship model depends on the middle layer.
If junior pathways collapse, the next generation of editors and format inventors never matures.
That risk is not theoretical. It is structural.
The Provenance Question
In an AI-saturated ecosystem, readers, schools, and institutions will increasingly ask:
Who wrote this?
What are their credentials?
Is this fact-checked?
Is this human-led?
Is it traceable?
Where are the bigraphies?
This is especially urgent in:
Children’s publishing
Educational publishing
Historical non-fiction
Cultural commentary
When authorship is blurred, trust erodes.
In the age of AI, author biography becomes part of the quality signal.
Books without visible provenance may increasingly feel unanchored.
Two Futures for Publishing
Scenario A: Industrial Acceleration
AI-led production pipelines
Reduced creative staffing
Emphasis on IP monetisation
Author identity secondary to brand
Rapid content cycles
This model increases margin.
But risks flattening originality.
Scenario B: Provenance Renaissance
Visible author biographies
Transparent creative credit
Human-led conceptual systems
Curated, traceable works
AI used as assistant, not origin
This model builds trust.
And trust becomes premium in noisy markets.
The Economics Behind the Anxiety
Economic research suggests AI’s impact depends on whether it complements or substitutes labour.
If it substitutes: wages compress, mid-tier roles disappear.
If it complements: productivity increases while human judgment remains central.
The direction is not technologically predetermined.It is shaped by corporate incentives and policy.
The Industrial Revolution did not automatically benefit workers; gains took decades to distribute. Publishing may face a similar lag.
What This Means for Creatives
Creatives who rely on:
Repeatable styles
Derivative formats
Task-based output
…are most exposed.
Creatives who provide:
Structural invention
Conceptual architecture
Visual systems
Cultural judgment
Taste leadership
…become more valuable.
AI can generate pages.
It cannot yet design enduring narrative systems.
Publishing Is Vast - So the Outcomes Will Be Uneven
Educational publishing may see rapid automation.
Trade fiction may fragment into AI-assisted genre mills and high-authority literary brands.
Children’s non-fiction may divide between fast algorithmic products and curated, research-driven work.
Academic publishing may integrate AI cautiously under peer review safeguards.
Indie publishing may explode with hybrid human-AI experimentation.
There will not be one future.There will be many.
This Is Not the End of Publishing
It is the end of invisible labour.
The question is whether publishing:
Doubles down on asset extractionor
Reasserts the value of human authorship
AI is a tool.
Whether it becomes an assistant or a replacement depends less on silicon and more on leadership.
Publishing has always survived technological shifts - from the steam press to desktop publishing to the internet.
The difference this time?
The tool now writes.
That changes the stakes.
Action
If you are:
An author
An illustrator
An editor
A rights professional
A teacher
A librarian
Ask one simple question:
Where does the human sit in your publishing model?
The answer will determine who thrives in the next decade.
We are told that every generation believes its crisis is unprecedented.
The mechanised loom terrified weavers.The steam press terrified printers.The paperback terrified hardback houses.Desktop publishing terrified typesetters.
The internet terrified everyone.
And yet books persisted.
But Artificial Intelligence is not a faster press. It is not a cheaper format. It is not a new distribution channel.
It is a system capable of imitation at scale.
And imitation, if left unexamined, has a curious way of becoming authority.
If publishing chooses to treat books as infinitely replicable units - optimised, templated, generic, accelerated - AI will oblige enthusiastically. It will fill the pipeline with remarkable efficiency.
But if publishing remembers that its true task is not production but discernment - not volume but vision - then AI may yet serve rather than supplant.
The danger is not that machines write.
The danger is that no one notices who no longer does.
For in the end, civilisation is not measured by how much content it produces, but by how carefully it decides what deserves to endure.
One hopes - perhaps unfashionably - that the guardians of books will
choose endurance over expedience.
But hope, like a first edition, is only valuable if preserved.
David Salariya is an architect of ideas - a publishing founder, format inventor and visual storyteller whose work has shaped children’s non-fiction for nearly five decades.
As founder of The Salariya Book Company, he built internationally successful series translated into over 35 languages, designing not just books but repeatable creative systems - visual grammars, narrative engines and the imprint identities of Book House, Scribblers, and Book House that travelled globally. His work blends conceptual architecture with accessible storytelling, making complex subjects engaging without flattening their depth.
He is currently working on Thirty-Two Books That Changed Britain, a cultural exploration structured around the physical logic of printing itself - examining how key texts shaped Britain’s intellectual, political and social identity. Alongside this, he writes and speaks about publishing, provenance and the future of creative work in an AI-saturated world.