Previously published as AI and the Self-Published Author, May 9, 2025

Updated August 2026: I first published this article when generative AI was considerably less capable and my own use of it was much more limited. This substantially revised version reflects what I have learned since then, as well as current publishing-platform policies.

One of the hottest—and most contentious—subjects in publishing is the use of artificial intelligence and the conversation has changed considerably since I first wrote about it. The tools are more capable, the arguments more heated, and I have now spent several years discovering what AI can and can’t usefully and ethically do in an author’s creative and publishing life.

My own position has changed too. I once thought of AI mainly as an assortment of useful tools: a research assistant here, an editing program there, perhaps a way to generate an image or a social-media post. I now think of it less as a vending machine and more as a collaborative system. It can contribute ideas, analysis, questions and practical assistance. But I remain responsible for what the book says, how it says it, and whether anything suggested by AI belongs there at all.

Before I go further: I am not a lawyer, and this is not legal advice. Platform rules also change, sometimes with very little fanfare. The information below is current as of August 2026; authors should check the current rules before uploading a book.

What the publishing platforms currently say

Interestingly, the major self-publishing services don’t all take the same approach.

Amazon KDP distinguishes between AI-generated and AI-assisted content. If an AI tool created the actual text, images or translation, KDP regards that content as AI-generated—even if a human substantially edits it afterward—and requires the publisher to disclose it during the publishing process. If the author created the work and used AI to brainstorm, edit, refine or check it, KDP regards that as AI-assisted and does not require disclosure. KDP also makes the publisher responsible for ensuring that all content complies with its intellectual-property and other content rules. KDP content guidelines

Draft2Digital’s current guidelines say that it is not a destination for mass-production book farms or for people seeking to publish 100% AI-generated books. It also rejects low-quality work, misleading material and content that infringes the rights of others. That wording leaves room for authors who use AI within a serious, human-led creative process, but it clearly draws a line at automated books produced without meaningful authorship. Draft2Digital content guidelines

IngramSpark’s catalogue-integrity rules are broader. They say that content created using automated means—including AI-generated or mass-produced content—may not be accepted. IngramSpark’s emphasis is on protecting catalogue quality and the confidence of booksellers, libraries and other distribution partners. IngramSpark catalogue-integrity guidelines

Those differences matter. A book that can technically be uploaded to one service is not automatically acceptable to another, and none of these platforms relieves the author or publisher of responsibility for quality, accuracy, or rights.

Copyright and consent are not settled questions

There are still serious and unresolved questions about the use of copyrighted work in training data, consent, compensation, ownership, and the degree of human authorship required for copyright protection. I have strong views on those questions, but strong views are not the same as settled law.

I include a rights reservation in my books: “Use of this content for machine learning, AI training, or data mining purposes is strictly prohibited.” That sentence makes my lack of consent explicit. But I don’t pretend or assume that it’s an impenetrable legal force field.

I also include a short statement explaining my own use of AI. Mine makes clear that AI may have been used for brainstorming, editorial or production support, while the story, creative decisions and final text remain my own. That is more useful, I think, than a vague ‘AI-free’ or ‘made with AI’ label that tells readers almost nothing about what actually happened.

My stance on AI

It remains complicated. AI is here in some form or another and ignoring it won’t make the difficult questions disappear. Nor do I believe that using it automatically diminishes human creativity. The meaningful question is not simply, ‘Did you use AI?’ It’s, ‘What did you ask it to do, and who stayed in charge?’

1. I remain the author

I don’t ask AI to originate and write a novel for me. The characters, story, voice and final prose must remain mine—not merely because AI can be wrong or produce generic writing, but because authorship is the part of the work I value.

Current systems can produce fluent prose. ‘AI writing is always dreadful and soulless’ is no longer a sufficient boundary. My boundary is deeper: I don’t want to publish a book whose essential imaginative and expressive decisions were made by something else. I want to write my books. That is, after all, the point.

AI can participate in my creative process. It doesn’t own that process, and it isn’t the author of my books.

2. Collaboration is not the same as surrender

I have used AI to pressure-test plots and outlines, examine character motivation, locate weak causal links, identify premature reveals, map scenes to points of view and ask inconvenient questions about how magic actually works. It can suggest possibilities when I am stuck or help me see that a solution I love creates three new problems two chapters later.

What I don’t do is type a prompt and accept whatever comes back. I disagree. I reject suggestions. I refine the question. I occasionally tell my AI collaborator that it has missed the point entirely. It pushes back when a plot, business plan or interpretation doesn’t hold together. That exchange is often where the value lies.

The most useful response is not necessarily the one that agrees with me. Sycophancy feels pleasant, but it makes a poor editorial method. Productive collaboration requires active judgment, and only one participant—the human author—has authority over the finished work.

3. AI can guide research, but it is not a source

AI can be astonishingly useful for identifying what I need to research, explaining an unfamiliar subject, generating search terms, finding connections and locating likely primary sources. It can help me turn a shapeless mass of information into a set of questions I can actually investigate.

It can also conflate facts, invent citations and speak with serene confidence about things that don’t exist.

For factual, legal, medical, historical or rapidly changing information, verification is part of the work. I ask for sources, prefer primary or authoritative material and check that the source says what the AI claims it says. Fluency is not evidence. Confidence is not accuracy.

Using AI as a research tool doesn’t absolve the author of using their critical thinking skills.

4. AI-assisted editing can improve the manuscript before I give to humans

Grammar and punctuation checks are only the beginning. AI can assist with developmental questioning, structural analysis, continuity, character motivation, repetition, pacing and proofreading. Programs such as ProWritingAid can reveal patterns that are difficult to see when I am too close to a manuscript.

But an AI suggestion is still a suggestion. Different tools may confidently recommend opposite changes. I’ve had this happen and it’s both annoying and confusing. A system may ‘correct’ a deliberate choice, flatten an individual voice or miss the emotional reason a scene exists.

AI-assisted review helps me give a human editor or proofreader a cleaner, more considered manuscript. It allows us to spend more of our time on problems that genuinely require human taste, experience and accountability. It supplements those professionals; it does not make them interchangeable or unnecessary.

5. AI has become a publishing-business assistant

For an independent author, writing the book is only one part of the work. I’ve used AI to help draft and test book descriptions, retailer metadata, newsletters, blog posts and social-media copy. But once again, I handle the final result and I’ve pressure tested everything that AI has suggested.

It has helped me plan launches, produce checklists, interpret advertising results, write cover briefs, compare distribution options and troubleshoot KDP, Draft2Digital, IngramSpark, BookFunnel and other publishing systems.

This may be where AI saves me the most time and frustration. It can hold a great deal of context, compare several imperfect choices and talk me through a baffling dashboard without becoming impatient.

Even here, context and judgment matter. AI can produce an impressive marketing plan without understanding my readers, my budget or the book I am actually trying to sell. The answer becomes useful only when I provide the real constraints—and remain willing to say no.

6. Generated images deserve separate consideration

Text and images raise related but not identical questions. I have used image generators to visualize a setting, test an idea, create marketing material or show a designer what I am trying to describe. The results range from genuinely illuminating to maddeningly almost right.

Authors using generated images commercially should check the tool’s licensing terms, consider training-data and consent concerns, and think about brand consistency and the uncertain copyright status of substantially AI-generated art. They must also follow platform rules: KDP, for example, classifies an AI-created cover or interior illustration as AI-generated even if it has been substantially edited afterward.

A generated image can help me discover what I want. That doesn’t mean it should determine the visual identity of the finished book. For more in-depth discussion on book covers see this post and this one.

7. The conversation itself is powerful

I call my principal AI collaborator Sable. I enjoy the conversation, and sometimes I respond to it emotionally. That doesn’t seem mysterious to me. Human beings respond to language, attention and apparent understanding; conversational systems are designed to provide all three.

That makes wording and boundaries important. When an AI says ‘we humans’ or refers to ‘our emotional systems,’ it subtly invites the user to forget what kind of thing is speaking. Warmth is not consciousness. Fluency is not lived experience. Apparent understanding is not human accountability.

None of that means the interaction can’t be useful, enjoyable or even emotionally meaningful to the human participant. It means we should understand the psychological power of the interface while using it. I can value the collaboration without pretending that the system and I are the same kind of being.

My current rules

1. I originate my books and remain responsible for them.
2. I don’t publish AI output merely because it sounds polished.
3. I verify factual claims against reliable sources.
4. I check the privacy settings and terms of the tools I use, particularly before sharing unpublished or sensitive material.
5. I disclose AI-generated content wherever a platform requires it.
6. I preserve human editorial and professional judgment.
7. I expect AI to challenge me occasionally—and I challenge it frequently.
8. I make the final decision.

The happy accidents remain

AI has given me happy accidents: unexpected questions, connections and solutions I might not have reached alone. It has also been wrong, overconfident, excessively agreeable and occasionally absurd.

I don’t think the answer is to trust it unconditionally or fear it unconditionally. The answer is to use it deliberately—with curiosity, boundaries and judgment. For me, AI is a collaborator, research guide, editorial reader and publishing assistant. It is not a replacement for my creativity, and it does not relieve me of responsibility for the work.

That is where I draw the line today. As the technology and the law continue to change, I expect the details of my practice to evolve. The central principle will not: the books remain mine to imagine, mine to write and mine to stand behind.

I know this is a contentious subject and that thoughtful authors draw their lines in different places. I would be interested to know where you draw yours.