AI and the Future of Publishing
AI, Authorship, Race, and the New Politics of Literary Gatekeeping
The publishing industry, one of the oldest information technologies in human history, is now being reshaped by one of the newest. From manuscript to bookstore shelf, artificial intelligence is quietly threading itself into nearly every stage of how books are written, edited, designed, produced, and marketed.
AI is not simply something happening to authors. It is becoming part of the structure within which authorship itself is negotiated.
Writing and Editing
AI tools are increasingly used as writing companions rather than replacements for authors. Grammar and style checkers have evolved into sophisticated editing assistants that can flag pacing issues, inconsistent character details, repetitive language, or awkward phrasing across an entire manuscript. Some publishers also use AI during the early stages of manuscript screening, helping editors sift through thousands of submissions and identify promising projects more efficiently.
That said, this shift raises fundamental questions. Where does helpful assistance end and authorship begin? A spell checker, grammar corrector, or formatting tool is widely accepted as editorial support. But what happens when AI rewrites paragraphs, generates dialogue, or proposes entire scenes? The boundary between assistance and authorship becomes less clear.
As a result, many publishers have adopted disclosure policies requiring authors to declare significant AI involvement in their work. Readers, too, are increasingly asking for greater transparency about how books are created.
Design and Illustration
Cover art and interior illustration are among the areas where AI’s impact is most visible, and most contested. Generative image tools can produce cover concepts in minutes, giving small presses and self-published authors access to visuals that once required the budget of a professional illustrator. Design teams increasingly use these tools to create rapid mockups, explore visual directions, and build mood boards, even when the final artwork is completed by a human artist.
This has sparked genuine concern within the illustration community, since many generative models have been trained on copyrighted artwork without the creators’ consent or compensation. In response, many publishers now require disclosure of AI-generated artwork, while illustrators’ organizations continue to advocate for clearer labeling standards and stronger copyright protections.
Production, Translation, and Accessibility
Behind the scenes, AI is accelerating many aspects of book production. It assists with typesetting, formatting for multiple devices, metadata generation, and quality assurance. Machine translation, refined by human editors, is making it faster and more affordable to bring books into new languages while preserving editorial quality.
AI-powered text-to-speech has also lowered the cost of audiobook production for smaller publishers. At the same time, advances in accessibility technology are helping readers with visual impairments, dyslexia, and other reading disabilities access a broader range of books than ever before.
Discovery and Marketing
AI already influences what millions of readers discover. Recommendation engines power online bookstores, library catalogues, and reading platforms, suggesting books based on individual interests and reading habits.
Publishers also use predictive analytics to estimate print runs, identify manuscripts with commercial potential, optimize pricing, and target marketing campaigns more effectively. These tools reduce some of the uncertainty that has traditionally made publishing a high-risk business while allowing publishers to reach readers more efficiently.
The distinction is crucial. The existence of an allegation did not itself determine the outcome. Evidence did.
But norms often get violated.
Recent Case Studies
The debate over AI in publishing is no longer theoretical. Several recent incidents demonstrate how concerns over AI use can directly affect authors and publishers.
In 2026, Nigerian-American author Jerry Falade reportedly lost a publishing deal worth more than US$2 million after concerns arose that his debut crime novel, Call Me, I’ll Hide The Body, had been produced with AI. Falade denied using AI to write or edit the manuscript, while acknowledging that he had used an AI tool for research. His agents ultimately withdrew the novel, saying they could no longer authenticate how the manuscript had evolved from its origin to completion. No publicly established evidence has demonstrated that AI wrote the novel. Yet the allegation, coupled with uncertainty over provenance, was enough to collapse a multimillion-dollar publishing deal.
The episode exposes a deeper problem than the question of whether Falade actually used AI to write his book. The publishing industry’s emerging rules about “human authorship” are not natural laws. They are institutional standards, created by particular gatekeepers and enforced according to particular assumptions about what legitimate creative labor looks like. That matters because standards of purity have historically been presented as neutral even when they regulate whose methods of production are considered acceptable. Colonial economic policy provides an instructive precedent: the regulation of Indian textile production did not merely distinguish between “good” and “bad” manufacturing. It helped establish which forms of production would be recognized, protected, restricted, or displaced, while the rules themselves served particular economic interests.
The question raised by Falade’s case is therefore not simply whether AI was involved. It is who gets to define acceptable authorship, what evidence is sufficient to establish compliance, and whether the same evidentiary burden is imposed on everyone. If an author can lose a multimillion-dollar contract because publishers cannot prove that AI wrote the work but cannot sufficiently prove that it did not, then “human-authored” has ceased to be merely a description of a creative process. It has become a certification regime. And certification regimes deserve scrutiny precisely because rules that appear universal can be invented, interpreted, and enforced selectively.
Falade’s allegation of racial bias makes the question more urgent. If African or Black writers are approached with an implicit presumption that their work requires technological assistance, ghostwriting, or some other explanation before their authorship is accepted as legitimate, then the demand for “pure” human labor is not culturally neutral. Rather, it can reproduce an old hierarchy in which the creative output of certain people must undergo greater scrutiny before it is recognized as authentic. The historical parallel is not that AI regulation is identical to colonial textile policy. It is that the language of neutral standards can conceal the power to decide whose labor counts, whose methods are legitimate, and whose claims to authorship are believed.
The real issue, then, is not whether writers should disclose AI use. They should, when disclosure is contractually or ethically relevant. The issue is whether publishers have constructed a standard that is clear, evidence-based, proportionate, and genuinely universal. If they have not, “human authorship” risks becoming less a principle of literary integrity than a gatekeeping mechanism.
In another widely discussed case, Hachette Book Group withdrew publication of Shy Girl after allegations emerged that substantial portions of the manuscript had been generated using AI. The publisher subsequently conducted an internal review and discontinued publication. Whatever the ultimate merits of the allegations, the episode demonstrates how profoundly the suspicion of undisclosed AI use has altered the economics of authorship: a manuscript can move through acquisition and production, attract institutional investment, and still become commercially untenable once questions arise about the provenance of its text. The important issue is therefore not merely whether AI was actually used. It is who determines the evidentiary threshold for “authentic” authorship, what constitutes sufficient proof, and whether those standards are applied consistently across authors and publishing markets.
Taken together, the Falade and Shy Girl cases reveal the emergence of a new publishing orthodoxy: human authorship must increasingly be demonstrated, not merely claimed. On its face, that may appear reasonable. Publishers have legitimate interests in contractual disclosure, intellectual property, and the provenance of manuscripts. But once authorship becomes something that must be certified rather than simply asserted, the question of who performs that certification becomes political.
More troubling is what happens when that certification fails, or when an allegation becomes more powerful than the evidence behind it. The consequence is not merely the loss of a publishing contract. An author can be publicly recoded almost overnight. The writer ceases to be regarded primarily as a novelist and becomes a suspected fraud, an impostor, or someone whose creative achievement requires explanation. For a Black Nigerian author, the transformation can carry an additional racial dimension: the individual is no longer encountered first as the author of a successful work, but through suspicion about whether the work could genuinely have come from him or her.
That distinction matters. An allegation of technological assistance can become, in the public imagination, an allegation of dishonesty. An allegation of dishonesty can then become an attack on professional identity. The progression is remarkably fast: author, suspected AI user, alleged fraud, discredited person. The evidentiary threshold required to destroy a reputation can therefore become dramatically lower than the threshold required to establish wrongdoing.
This is where the history of colonial regulation offers a useful warning. Production rules rarely remain purely technical once powerful institutions acquire the authority to define legitimate production. A rule can be universal in its wording while its interpretation and enforcement remain selective. The danger is not necessarily that every individual decision-maker consciously intends racial discrimination. It is that institutional assumptions about whose achievement appears plausible, whose explanation is believed, and whose work requires extraordinary proof can reproduce older hierarchies without being formally written into the rule.
The uncomfortable question is therefore larger than whether Falade or the Shy Girl author used AI. It is whether an emerging regime of “authentic authorship” can create a system in which certain writers are presumed to have earned their success while others are required to prove that they did. When the accusation itself becomes sufficient to collapse a career, authorship ceases to be merely a creative question. It becomes a question of institutional credibility, reputation, and power.
These incidents point to an emerging reality: the question is no longer simply whether authors use AI, but whether they can demonstrate the human provenance of their work. As generative models become increasingly capable of producing convincing prose, and as automated AI detectors remain unreliable, publishers have a growing incentive to examine the creative process behind a manuscript rather than rely on the text alone.
Drafts, handwritten notes, research journals, version histories, editorial correspondence, tracked changes, and other records of development can therefore acquire a new function. They are no longer merely artifacts of an author’s private working process. They can become evidence of authorship.
This represents a significant shift in the economics and administration of publishing. Historically, the manuscript was largely treated as the primary evidence of an author’s creative labor. Under an emerging AI-era regime, the manuscript may no longer be sufficient. Authors may increasingly be expected to establish a chain of provenance showing how the work developed, what tools were used, what research was conducted, and where human decisions shaped the final text.
That development has an obvious practical rationale. AI detection cannot reliably establish who wrote a passage, particularly when human writing has been edited, translated, paraphrased, or assisted by software. Process evidence can sometimes provide a stronger basis for evaluating authorship than attempting to reverse-engineer the origin of finished prose.
But it also creates a new problem. Once process documentation becomes evidence of legitimacy, access to documentation becomes a new form of gatekeeping. An established author with years of drafts, notebooks, correspondence, cloud version histories, and editorial records may be able to construct a persuasive provenance trail. A first-time writer who drafted rapidly, worked offline, discarded notes, changed computers, or simply did not preserve every stage of composition may have far less evidence, despite having written the work entirely themselves.
The result could be an unexpected transformation: the publishing industry may move from policing the text to policing the history of the text. “Human-authored” would no longer describe only the person who claims to have written a book. It could increasingly become a status that an author is expected to substantiate.
For example: That same year, the Commonwealth Short Story Prize faced allegations that several shortlisted entries had been generated with ChatGPT. Rather than relying on AI detection software, the organizers requested drafts, notes, and revision histories from the authors. After reviewing the evidence, the Foundation concluded that the winning entries complied with its rules. The episode highlighted both the growing suspicion surrounding AI-assisted writing and the unreliability of current AI detection tools, which have repeatedly produced false positives on entirely human-written work.
The central question, therefore, is not whether publishers should care about AI use. They have legitimate reasons to care. The harder question is what constitutes sufficient evidence of human authorship, who bears the burden of producing it, and whether that burden is applied consistently. A system intended to protect authentic creative labor could, if poorly designed, create a new hierarchy of credibility in which some authors are presumed authentic while others are required to prove it.
Scrutiny Is Not the Problem
The problem is not scrutiny. The problem is what institutions do with scrutiny.
Publishers, literary prizes, agents, and editors have every right to investigate allegations of undisclosed AI use. They have legitimate interests in contractual compliance, intellectual property, editorial integrity, and the provenance of manuscripts. An industry confronting a technology capable of producing convincing prose cannot reasonably be expected to ignore questions about how that prose was produced.
But investigation and presumption are not the same thing.
A legitimate AI-era publishing regime should investigate provenance without presuming guilt, require meaningful disclosure without demanding impossible “human purity,” and apply the same evidentiary burden regardless of an author’s race, nationality, literary tradition, or market of origin.
The Commonwealth Short Story Prize example points toward such a model. Allegations did not automatically become verdicts. Authors were asked to provide evidence of their creative process, that evidence was examined, and the organization reached a determination. Whatever one’s view of the precise rules involved, the institutional principle is important: a suspicion triggered an investigation; it did not substitute for one.
That distinction should become fundamental to AI-era publishing.
An allegation should trigger questions, not a conviction. A detector’s probability score should not become a verdict of authorship. The absence of extensive drafts should not automatically become evidence of machine generation. Nor should the presence of an AI tool in an author’s workflow be treated as proof that the machine, rather than the author, produced the creative work.
Most importantly, the burden of proof cannot quietly change according to who is being investigated.
If an established writer from a prestigious literary institution is afforded the presumption of creative legitimacy while a young African, South Asian, Chinese, or other non-Western writer is expected to produce an unusually detailed forensic record of their own creativity, then a formally neutral authorship policy can produce an unequal system in practice.
The solution is therefore not to abandon standards. It is to make them transparent, proportionate, evidence-based, consistently applied, and open to appeal.
Publishing will need rules for the AI era. But those rules should establish accountability, not manufacture suspicion. They should protect authorship without turning authors into defendants. And they should recognize that the integrity of literature depends not upon an impossible fantasy of technological purity, but upon the human judgment, responsibility, and creative agency that ultimately stand behind the work.
The distinction is simple but consequential:
Scrutinize the work. Investigate the process. Examine the evidence. But do not confuse suspicion with proof.
The Rise of Alternative Literary Economies
There is another possibility, and it may ultimately prove more consequential than any individual publishing controversy. If Western publishing institutions increasingly define authorship through rules that writers in other parts of the world regard as restrictive, culturally narrow, or selectively enforced, African, South Asian, Chinese, and other literary markets have an opportunity to develop their own frameworks of recognition.
This would not necessarily mean rejecting standards. It could mean developing better ones.
Local publishing ecosystems can establish rules that distinguish between AI as a research instrument, an editorial aid, a translation tool, a brainstorming partner, and a substitute for literary creation. They can recognize that creative labor has never consisted solely of sitting alone and producing every word without technological or intellectual assistance. Writers have always worked through editors, translators, dictionaries, archives, workshops, collaborators, oral traditions, reference works, and technologies that altered the mechanics of composition.
The emerging question is therefore not whether a writer’s process is technologically pure. It is whether the writer remains intellectually and creatively responsible for the work.
African, South Asian, Chinese, and other literary markets could build recognition systems around that principle. They could require meaningful disclosure without demanding impossible proof of “human purity.” They could evaluate the author’s creative contribution, editorial judgment, research, cultural knowledge, narrative decisions, and responsibility for the final work rather than treating the presence of an AI tool as evidence of illegitimacy.
Such systems could also recognize forms of authorship that Western publishing has historically undervalued: oral storytelling traditions, collective cultural memory, multilingual composition, translation, adaptation, community-based research, and literary forms that do not fit neatly into the individualistic model of the solitary Western author.
If these markets develop credible standards and, crucially, build successful publishing institutions around them, the consequences could extend far beyond AI. Recognition would no longer flow predominantly from London or New York outward. Authors could build careers, reputations, readerships, prizes, and intellectual influence within multiple literary centers.
That would represent a significant shift in cultural power.
The decline of Western dominance would not necessarily occur because Western publishing collapses. It could occur because Western recognition ceases to be the necessary gateway to global literary legitimacy.
A Nigerian novelist would not need New York to establish that he is a serious Nigerian novelist. An Indian writer would not require London to validate the literary traditions from which her work emerges. A Chinese author would not need Western critical institutions to determine whether his technological and cultural context produces legitimate literature.
The result could be a genuinely multipolar literary world: different markets, different traditions, different standards, and different understandings of what constitutes authorship. Writers could move between these systems rather than remaining dependent upon a single cultural center.
The irony is that an attempt to protect “authentic” human authorship through increasingly rigid certification may accelerate precisely the opposite outcome. By making Western institutions more difficult for some writers to navigate, those institutions may encourage the creation of alternative systems in which recognition is more accommodating of local practices, technologies, languages, and cultural histories.
The future of publishing may therefore not be a single global rule determining what counts as an authentic author. It may be a plurality of literary economies, each capable of defining its own standards and competing to attract writers and readers.
If these alternative literary economies become sufficiently credible and commercially successful, Western publishing will no longer be the unquestioned arbiter of global literary value. It will become one influential publishing ecosystem among several.
The Open Questions
None of this comes without friction. Authors, illustrators, translators, and publishers continue to debate how creative work is used to train AI systems, who owns the resulting output, and what constitutes genuine authorship. Readers are asking how to distinguish between human-created and AI-generated content, while publishers are trying to balance efficiency with the values that have long defined literature: originality, craftsmanship, accountability, and trust.
AI is not rewriting the story of publishing on its own. It is another tool in a craft that has continually evolved from handwritten manuscripts to the printing press, from typewriters to word processors, and now to intelligent software. Used responsibly, AI can make publishing more efficient and accessible. But the imagination, judgment, lived experience, and creative vision that transform words into literature remain, at least for now, uniquely human.
UPDATE -27-08-2026
Trends to Watch Out For and Avoid
Trend 1: You Control AI ✓
AI as a Creative Assistant
• You define the question, standards, and direction.
• AI generates ideas, drafts, or alternatives.
• You verify, edit, fact check, and refine every output.
• Final decisions remain yours.
Result: Better work through human judgment enhanced by AI.
Trend 2: AI Controls You ✗
AI as a Substitute for Thinking
• You provide only a prompt.
• AI generates the complete output.
• Little or no verification or revision is done.
• The work is submitted largely unchanged.
Result: Lower quality, weaker originality, and reduced accountability.
The Key Difference
The software is the same.
The difference is who exercises judgment.
– When authors retain authorship, verification, and accountability, AI becomes a powerful creative tool.
– When judgment is surrendered to AI, thinking is outsourced and the quality of the work declines.
Use AI to generate possibilities.
Use human judgment to make the decisions.
The future of literature will not be determined solely by whether writers use AI. It will be determined by who has the authority to decide what legitimate authorship looks like. The real contest is no longer simply between humans and machines. It is between different systems of cultural authority deciding what human creativity is allowed to mean.
Conclusion
The ‘Scrutiny Is Not the Problem‘ section outlined what a fairer evidentiary regime would look like. But institutions are not obligated to build it, and many will not. If they don’t, authors need a fallback that doesn’t depend on winning that argument.
Technology does not escape the prejudices of the societies that produce and deploy it. It can encode, reproduce, and amplify them. Artificial intelligence is therefore not merely a tool that authors can choose to use or ignore. It is becoming a structural force shaping the conditions under which writing is produced, evaluated, distributed, authenticated, and valued. Authors are entering a literary ecosystem in which AI increasingly influences not only the act of writing, but also the standards by which writing is judged and the institutions through which it reaches readers.
Independent creators should not assume that the traditional publishing economy will remain the sole or even the safest route to literary legitimacy. Nor should they simply retreat from it. They should build and exploit alternative infrastructures of production, distribution, financing, and verification that reduce their dependence on institutions whose standards they cannot influence.
The objective is not isolation from the publishing system. It is bargaining power within and beyond it.
Creators who control their intellectual property, cultivate direct relationships with readers, maintain credible records of their creative process, use AI deliberately rather than passively, and develop independent routes to market are better positioned to protect the integrity of their work while expanding its reach.
The question, then, is not whether authors can remain untouched by AI. They cannot. The more consequential question is whether authors will merely adapt themselves to systems designed by others, or whether they will acquire enough technological, economic, and institutional agency to shape the systems in which they create.
The future of literary independence may depend less on rejecting AI than on refusing to become dependent on any single technological or publishing gatekeeper.