The AI author is already here. The publishing industry just hasn’t decided how to cope with it.

About every technological shift in publishing has been accompanied by a moral panic. The rotary press supposedly cheapened literature. Word processors were accused of making writing mechanical. Search engines would destroy research. Today, artificial intelligence occupies that familiar role. It has become the latest proxy in a much older debate: what exactly do we value when we buy, publish or read a book?

I have no hesitation in stating my own position. I use AI extensively in several functions concerning my own writing. I also reject a great deal of what it produces.

Those two statements are perfectly compatible.

Like every experienced writer, I judge a tool by its output rather than by its existence. A fountain pen, a typewriter, Microsoft Word, OCR software, grammar checkers, search engines and translation tools all changed the writing process. None of them replaced the responsibility of the author. Large language models belong to the same continuum, although they are undeniably more powerful.

Ironically, “AI-free writing” has already become almost impossible. AI quietly assists spelling, grammar, search, transcription, translation, image generation, note taking, indexing and research in countless applications. Anyone wishing to avoid AI completely would have to retreat to a log cabin with a mechanical typewriter – and probably give up the internet altogether.

Yet many authors publicly denounce AI. Some genuinely dislike it. Others privately use it while condemning it in public. The ambiguity is understandable because publishers themselves send mixed messages.

On the one hand, publishers emphasise the authenticity of the author’s voice. Dust jackets celebrate personal experience, unique perspectives and genuine expertise. Readers are assured that these are the author’s own words and thoughts.

On the other hand, commercial publishing has long accepted practices that complicate this narrative. Celebrity memoirs, management books and political books frequently rely on ghostwriters or substantial editorial intervention. Particularly in books expected to reach bestseller lists, the “author” is often a name giver but otherwise remote participant in a much larger production process. Nobody seems particularly troubled by this. Apparently, invisible human collaborators remain acceptable, while invisible digital ones provoke outrage.

The distinction is becoming increasingly difficult to defend.

The industry’s public discomfort contrasts sharply with its private experimentation. Publishers across academic and professional markets have begun issuing practical guidance rather than outright prohibitions. Wiley, for example, published detailed AI guidelines for authors in 2025, explicitly recognising that many writers want responsible ways of integrating these tools into their workflow while preserving accuracy, originality and their own voice.

That approach strikes me as considerably healthier than pretending the technology does not exist.

Economic realities reinforce this shift.

During the past generation, earning a living solely by writing books has become markedly harder. Only a small minority of authors can afford to spend one or two years researching and writing every manuscript. Bestseller authors, celebrities, recognised experts and personalities with enormous online audiences remain exceptions. Everyone else faces shrinking advances, growing marketing expectations and increasing commercial uncertainty.

Professional authorship has therefore become less sustainable than it once was. Many talented writers simply cannot afford to remain full-time authors.

Against that background, AI is less a luxury than an economic instrument.

Used intelligently, it accelerates background research, organises notes, compares sources, produces first drafts of routine passages and identifies structural weaknesses. None of these activities constitutes finished writing. They simply reduce time spent on repetitive tasks, allowing authors to devote more energy to the genuinely creative work that still requires human judgement.

The same commercial pressures have changed publishing itself.

For acquiring editors, the decisive question is often no longer whether a manuscript demonstrates extraordinary research or exceptional prose. Those qualities still matter but they rarely come first. Increasingly, editors ask different questions. Does the proposal fit our programme? Does it address a recognisable audience? Can we market it successfully? Does the author already possess a platform?

Publishing has always balanced editorial judgement with commercial calculation. Today the balance has shifted further towards marketability.

Speed has become another decisive factor.

In non-fiction, being first often matters as much as being comprehensive. Political developments, technological breakthroughs, business trends and cultural debates have increasingly short commercial windows. Fiction is beginning to follow the same pattern in several genres, particularly where readers expect frequent new releases rather than singular literary events.

This is precisely where AI delivers measurable advantages.

Self-publishing demonstrates the point particularly well.

Independent category-fiction authors frequently publish several books each year without advances, large print runs or guaranteed sales. Every manuscript represents a financial gamble. AI-assisted research, editing and drafting reduce production costs and shorten publication cycles, thereby lowering economic risk. Whether traditional publishers like the development is largely irrelevant. The ecosystem already exists and it continues to grow.

Traditional publishers may eventually adopt some of these production methods in genres where speed matters more than literary authenticity or where readers simply care more about receiving the next instalment quickly than about how every sentence reached the page.

This possibility makes one recurring argument especially problematic.

Many publishers have long assumed they could reliably distinguish human writing from AI writing.

In practice, they cannot.

Current language models rarely produce brilliant prose. They usually produce competent, average prose. Unfortunately, average prose has never been unusual in publishing.

Detection tools fare no better. They typically return probabilities rather than definitive conclusions. False positives remain possible, while determined users can often evade detection through revision. Recent controversies surrounding the horror novel “Shy Girl”, whose publication was disrupted after allegations of undisclosed AI involvement, demonstrated not only the reputational risks but also the remarkable uncertainty surrounding available detection methods. Even multiple AI detectors produce probabilistic rather than conclusive assessments.

That uncertainty creates obvious contractual and legal difficulties. Publishers may suspect AI assistance without possessing evidence capable of supporting serious editorial or legal action.

Readers present another paradox.