Against AI Prose

In C.S. Lewis’s Till We Have Faces, Queen Orual writes the story of her life to indict the gods. “Being, for all these reasons, free from fear, I will write in this book what no one who has happiness would dare to write. I accuse the gods, especially the god who lives on the Grey Mountain.” The gods have been unjust, Orual believes; the story of her life is her complaint. In Part Two of the novel, Orual presents her case before the gods and receives their just judgement. As she does so, all of her self-righteousness falls away; the reader realizes that Orual’s hatred is its own indictment. Articulating her complaint with nothing left to hide behind allows Orual to see herself most clearly.

Writing is a process whereby we learn what we have to say and to say it clearly. Till We Have Faces is Lewis’s final novel and the one he called his best book. It is a beautiful story. It is also an excellent depiction of what writing is really all about. Fr. Andrew Lazo argues that Lewis is always writing his autobiography—he did so in “Early Prose Joy,” most self-consciously in Surprised by Joy, and, Lazo’s argument goes, most clearly in Till We Have Faces. Writing, Lewis illustrates, is a process whereby we learn what we have to say and to say it clearly. It is a revelation of the self; it stands upon all previously read and written material. It is through writing that we speak into the literary conversation that transcends particular times and places.

Both the printing press and the internet extended human work. Generative AI replaces it. Writing faces a new threat in 2026. AI has gotten better; it can generate wordcounts previously unthinkable. And, while it still hallucinates, generative AI can pull quotes, format footnotes, and write copy in a far more impressive way than it could just a year ago. University systems are rushing to integrate AI into all aspects of higher education: The Martin Center’s Shannon Watkins recently published an interview highlighting UNC-Chapel Hill’s approach. Thales College (my institution) illustrates a different path: We held a Faculty Roundtable on AI in education and then spent two months crafting an AI policy limiting its use to protect classroom integrity.

There is obviously a need to contemplate the nature of AI as technology. In Against the Machine, Paul Kingsnorth argues that AI is shaped to replace human work. That, Kingsnorth argues, is what makes AI different from the printing press or the internet; both of these revolutionary technologies extended human work. Generative AI replaces it.

Kingsnorth has since launched the Writers Against AI campaign. Kingsnorth argues that there is something inescapably human about writing, and he invites those who share his view to affirm three commitments. If they do so, those who want to join the Writers Against AI campaign can use these easily downloadable graphics. The three commitments of the campaign are:

• I will not use AI in my work as a writer.

• I will not support writers who use AI in their work.

• I will support writers, illustrators, editors and others in related fields whose work is entirely human-made.

These are a helpful, limited set of propositions. They are not seeking to condemn or ban AI outright, nor are they denying the potential efficiency gains to be had in sorting information, in representing data, or in other specific use cases. Instead, these propositions seek to look at writing as something that merits positive defense.

Early adopters and those who are building AI tools see their work as revolutionizing the world. There remains a brief moment at the early stages of technological innovation when contemplation can prevent rash implementation; we are still at such a moment. We could still stop, pause, and ask, “Is this a task for which AI ought to be used? What is the tradeoff for handing this task to AI?” The time to question tech is at the beginning; we still have that possibility with AI.

Writing is personal. It is the expression of ideas communicated from one person to other persons. In “Mythopoeia,” Tolkien describes writing as “living images that move from mind to mind.” Writing is not just generating words; it is cultivating the expression of thought such that the ideas move from the original mind to the minds of the readers. To outsource part of the process would be to first diminish thought and then to remove from the reader substantive ideas. Writing is irreducibly personal.

To outsource part of the writing process would be to first diminish thought and then to remove from the reader substantive ideas. If writing is personal, then it fits within what Eastern University’s Brian Williams calls “an economy of beatitude,” in which persons give gifts. In a panel discussion on AI and classical education, Williams contrasted two views on writing within a school setting. Under an “economy of production,” the goal is to produce as many words as possible. Under such a telos, using AI to develop wordcount and edit its own process makes for an efficient use of time. Williams contrasts this perspective with what he calls “an economy of beatitude.” In this view, writing is a gift that one person crafts over time. The goal is not quantity of words but, rather, crafting the right expression.

Using AI to do any part of the process introduces a level of dishonesty that disrupts it. The process of teaching writing, and of becoming a writer, involves moving deeper and deeper into the economy of beatitude through cutting words; rearranging sentences, paragraphs, and evidence; and rewriting to produce the best expression possible. It is done, ideally, as an apprenticeship in which an older, skilled writer can deconstruct student writing and help the student piece it back together. In the process, the student learns the essential habits of wordsmithing. Helping students enter the economy of beatitude is the goal of academic writing. His writing is the gift that the writer offers to other persons.

My final argument against using AI in writing is directly related to the kind of writing I do: book reviews and published humanities research. Since 2015, I’ve written between five and 15 book reviews a year. My goal in a book review is to help a reader determine if a book is worth his time; typically, the review is a springboard into some larger cultural, historical conversation. My value as a reviewer lies in two elements: I have, in fact, read the book, and I have an argument to make about it that I can demonstrate. AI could no doubt offer a summary of a book and enable me to appear to have read it; I suspect that, over time, it could even become quite skilled at pulling out quotes along key themes I identified.

Any review I wrote using such methods would be a lie. I might be able to increase my output in terms of reviews published; I can even imagine a world in which I could have AI generate the skeleton and then the body of a review, while I still managed to tweak it to sound human. Such a review would be worthless, because I would have robbed myself of the ability to truly stand for or against a book.

The same is true of my published research. At this point, I have three published book chapters, a book (Sons of Adam, Daughters of Eve: C.S. Lewis’s Images of Gender), and three future book projects outlined. My method is a very traditional one for humanities research: I read dense, old books and write accessible criticism that helps other people get into those topics. In all these projects, my ability to write honestly derives first from doing my own research and then from crafting that research into a particular form that communicates with readers. Using AI to do any part of that process introduces a level of dishonesty that disrupts it.

If Lewis is right, then writing is formative, shaping the writer’s voice and helping him say what he has to say well. The whole process—ideation, research, outline, drafting, editing, publication—involves developing a series of ideas and arranging those ideas to express an argument in a way that is persuasive to others. Using AI for any of these tasks replaces human effort and reduces the human communication inherent in writing. For writing to be formative, it must remain human. Without effort, without the intentional rejection of AI in this specific domain, human writing will become filled with just as much AI slop as any other part of the internet.

To riff on that great Old Testament Joshua, “As for me and my writing, we will remain human.”

Josh Herring is professor of classical education and humanities at Thales College and the author of Sons of Adam, Daughters of Eve: C.S. Lewis’s Images of Gender (Davenant Press, 2026).