Stephen Marche: Writers Should Accept AI, But the Value of Human Text Doesn't Disappear

Language that sounds convincing but has no connection to reality has become the main byproduct of the generative AI era—and that's precisely why the role of the living author doesn't disappear, but transforms. Canadian writer Stephen Marche, who himself experimented with writing a novel with AI's help, believes: artificial intelligence has already permanently entered the literary and media process. But instead of ending the profession, it brings a more unpleasant shift—it sharply devalues template-like, averaged style and forces authors to prove their worth not through polished phrases, but through content, accuracy, and experience.

Marche begins with an everyday scene at a playground. During an argument, a girl shouts to a boy: "That's AI!"—meaning a new type of nonsense—speech that sounds meaningful but in fact has no connection to reality.

For the author, this is an important signal: children have quickly learned to recognize what adults are still trying to describe through terms like "hallucinations" or "synthetic content." AI has already become a cultural category, not just a technological tool. If previously bad text could disguise itself as normal, now readers are developing a new instinct—distrust of text that's too smooth, impersonal, suspiciously convenient.

According to Marche, treating AI as an apocalypse or, conversely, as a universal solution to all problems makes no sense. It's not the end of language or a magical replacement for creativity, but a powerful destructive tool that changes the economics of writing. Against this backdrop, the scandal surrounding Mia Ballard's novel Shy Girl is telling.

Hachette Publishing canceled the book's publication after accusations that the text relied on AI generation. Meanwhile, the novel had already been released as a self-published edition, and neither readers nor editors, judging by the reaction, saw a problem until AI use was explicitly named. Ballard herself claimed that an acquaintance who had edited an early version used AI, not her.

This episode essentially exposes the main nerve of the debate: society hasn't yet developed stable rules for what counts as impermissible use of AI in literature and where the boundary lies between editing, co-authorship, and complete generation. But something else is already clear: command of banal style is ceasing to be a scarce skill. Machines are capable of rapidly producing smooth texts, standard descriptions, safe introductory paragraphs, and endless variations on familiar themes.

Therefore, the value of what is harder to automate grows: authorial perspective, personal responsibility for meaning, the ability to connect words with reality, and form with observation, risk, and concrete human experience. A good writer is now distinguished not by the ability to write without errors and clichés, but by the ability to say something true. For the industry, this means a reassembly of familiar quality criteria.

Editors, publishers, and readers will have to rely less on external literariness and look more carefully at the origin of the text, its internal logic, and the degree of authorial participation. For authors themselves, the task also changes: AI can accelerate draft work, the search for formulations, or structuring of material, but it does not remove responsibility for truth, nuance, and intonation. The easier it becomes to produce plausible text, the more expensive becomes text backed by a real person and recognizable thinking.

Marche's conclusion is simple: we will have to accept AI because it's already embedded in the culture of writing. But we don't need to capitulate to it. In a world where machine speech increasingly imitates meaning, an author's value is determined not by the ability to generate volume, but by the ability to distinguish living experience from verbal noise.

And that is what makes writers not less, but more important.