The biggest sin in using AI in the arts right now is dishonesty.
It is not simply using the technology. I use it. I have been open about using it. I have explained my process in my books and on my blog, and I invite people to look closely at what I have done. Decide whether the finished work has merit. Decide whether it contains thought, experience, revision, and care—or whether it is just another piece of AI slop.
I understand why people devalue AI-assisted work. We are drowning in material that was generated quickly, barely examined, and released by people who seem more interested in exploiting a new market than making anything worthwhile. AI did not invent that kind of opportunism. Cheap books, B-movies, exploitation films, disposable pornography, formulaic music, and imitative art existed long before anyone could type a prompt and receive a finished product. Some of that work found strange and lasting value. A great deal of it was always slop.
AI did not create the opportunist. It gave the opportunist an assembly line.
What Is AI Slop?
To me, AI slop is not simply anything made with AI. It is work produced without enough human attention, judgment, revision, or responsibility.
It may look correct. It may be grammatically clean. It may even be impressive for a few seconds. But nobody stayed with it long enough to ask whether it was honest, coherent, necessary, or alive. The creator generated it, accepted it, and released it.
You can see this easily in AI book covers. Many are too rich. Every dial has been turned all the way up: dramatic lighting, glowing particles, perfect faces, ornate costumes, smoke, fire, impossible architecture, and elaborate lettering. The image initially looks expensive, but nothing has been restrained, so nothing feels important. It contains decoration without design.
Music may be an even more troubling example. A complete recording can involve a songwriter, performers, studio musicians, an arranger, engineers, and a producer. Generative AI can imitate the combined result of all that labor and create thousands of tracks almost instantly. Sampling and synthesizers also transformed music, but they remained instruments that required someone to decide what fragment mattered, how it should be changed, and what new work should surround it. A music generator can now offer the voice, lyrics, arrangement, instrumentation, performance, and production at once.
The issue is not just whether the result sounds good. It is whose identity and labor made that result possible, whether they consented, and whether the person releasing it is being honest about what happened.
The Book Problem
On the book side, the deepest ethical problem for me is that authors’ work has been put into AI systems without their knowledge or permission. Writers may have spent years researching and shaping a book only to see it absorbed into a system that can imitate, summarize, or compete with them.
That issue is not comfortably distant from me.
While working on my Seen / Unseen Greenville series, I put material from an older book about Greenville into AI. At the time, I was trying to recover historical information that was difficult to find elsewhere. I did not fully consider the ethical question of placing that author’s work into the system. I also did not credit her as clearly as I should have in the writing that followed.
Part of me can defend the decision as preservation. The book contained local history that might otherwise remain inaccessible or disappear. The project itself was intended to examine the historical pressures underneath Greenville’s neighborhoods—the things a growing city remembers, conceals, displaces, and forgets. Recovering that information had value.
But preservation does not erase authorship. Historical facts and an author’s expression are not the same thing. Even when information deserves to survive, the person who gathered and interpreted it deserves acknowledgment. The author is no longer in a position to grant permission herself, and I hope eventually to speak with her family, learn more about her work in Greenville’s historical community, and properly recognize what she contributed.
That does not undo the original decision. It does mean I can take responsibility for it.
This is why simple declarations about being “pro-AI” or “anti-AI” are not enough. The ethical questions change according to what was used, who created it, what permission existed, how substantially it shaped the result, and whether the final author tells the truth about the process.
Then Hugh Howey Released NEO
In August 2026, Wool author Hugh Howey released NEO, a free, distraction-free application designed specifically for writing books. It runs locally, requires no account or subscription, and keeps works in progress as readable files on the writer’s own computer. Its source code is publicly available.
NEO is deliberately simple. It includes an outlining area, automatic chapter numbering, placeholders for facts or problems that need to be addressed later, and a “Darlings” section for passages removed from the manuscript but not discarded. Writers can set daily goals, run word sprints, enter a more restrictive distraction-free mode, and export their work into several common formats.
What interests me most is what the current version does not appear to contain: an AI system waiting to complete the next sentence.
There is no ever-present invitation to generate a description, rewrite a paragraph, plan the next scene, or make the prose more dramatic. When you do not know a fact or cannot solve something immediately, NEO lets you leave a placeholder and continue. The unresolved problem remains yours.
Howey also included a feature that allows writers to email themselves timestamped PDF snapshots accompanied by a SHA-256 fingerprint. A digital fingerprint cannot prove that a human wrote the words, but it can help document that particular versions existed at particular times and were not subsequently altered. Howey has connected the idea to a growing problem: as AI becomes capable of generating not only manuscripts but simulated revision histories, writers may increasingly want some record of a work developing over time.
NEO is still very early software. Howey himself described the initial release as not even ready for beta, and users have reported installation and formatting problems. It should not be treated as a polished answer to every writer’s needs. But its limitations may be part of what makes it interesting. It is trying to create a room for writing rather than a machine that promises to write for you.
Drawing a Boundary
I want to try an experiment with my next book.
I want to develop the concept, research notes, characters, themes, and chapter outline in advance. Some of that planning may include conversations with AI. I will not pretend otherwise. Once that material is assembled, however, I want to freeze it. I want to carry the outline into NEO and write the entire first draft without AI-generated prose or further AI inspiration.
If I cannot remember a name, I can leave a placeholder. If a transition does not work, I can mark it and keep going. If I cannot solve a scene, I can sit with it until I do. The point is not to make writing artificially difficult. The point is to discover what happens when I establish a clear boundary and protect the space on the other side of it.
I could then describe the process honestly:
> This book was planned through AI-assisted discussion and outlining. Every sentence of the first draft was written by me without generative assistance. Any assistance used during later revisions is identified separately.
That statement is more meaningful than calling the book either “AI-written” or “AI-free.” The outline would influence the manuscript, so claiming complete freedom from AI would be false. But planning assistance is also not the same as asking a system to manufacture the sentences and scenes.
I am curious about what my voice will sound like in that space. Not because I believe AI has erased my voice from my previous work, but because I want evidence rather than anxiety. I want to compare a deliberately human-drafted manuscript with the writing I have developed through a more collaborative process. I may discover weaknesses. I may discover abilities I have not trusted enough. Either result would teach me something.
Human Writing Should Include Humans
Writing without AI-generated prose does not mean writing without help.
My goal has always been to include more people in the process. If I had more money for professional editing, I would use it. I would like to work with beta readers who can tell me where they became engaged, where they became confused, and where the language felt too polished or distant from me. That process has been difficult because sharing autobiographical work can make criticism feel like a judgment of the life behind it. Confidence and mental health are real barriers, even when collaboration is something I value.
But the answer is not necessarily isolation. It may be a smaller and safer human process: one chapter, one trusted reader, and a few specific questions at a time. Human authors have always relied on editors, readers, researchers, conversations, and other writers. Receiving help is not the same as surrendering authorship.
The meaningful distinction is responsibility. Who made the decisions? Who examined the language? Who rejected what did not belong? Who remained with the work when the easy answer was not good enough? Who finally stood behind it?
Is There a Place for AI-Assisted Writing?
I believe there is a vacuum in the market for a place that collects worthwhile, transparently AI-assisted writing. But it cannot simply be an unrestricted dumping ground for generated books. That would become an ocean of slop almost immediately.
What is missing is closer to a curated literary journal or selective press—one where every work includes a clear account of its process. Readers could see whether AI was used for brainstorming, outlining, research organization, developmental discussion, sentence-level editing, generated passages, or most of the manuscript. The work would still have to survive human editorial review.
The promise would not be that AI makes literature better. The promise would be that nobody is hiding how the literature was made.
The current reaction against AI may make such a project difficult. It may also be precisely why one is needed. Readers are being asked to choose between blanket rejection and blind acceptance when many of the real questions exist between those extremes. They deserve enough information to decide what kinds of assistance they will accept.
No Place—or the Right Place?
I do not think AI has no place in writing. I think it needs a defined place.
It can help someone organize years of memories, challenge an idea, locate a weakness, or see a structure that was previously hidden. It can also imitate an author, exploit uncredited labor, flatten a voice, and allow someone to publish work they have barely read. The technology does not erase the difference between those acts.
NEO appeals to me because it offers something increasingly rare: an empty page that does not immediately volunteer to fill itself.
My experiment will not prove that my work is purer than anyone else’s. It will not resolve every ethical question about AI, including the questions raised by the planning stage itself. It will simply establish an honest boundary: AI may help me draw the map, but I will walk the ground and write down what I find.
There may be a place for AI in the process. There must also remain a place where it stops.
Frank M. Anderson's Writing 
Leave a comment