I founded Page & Podium Press on the very human premise that our lived experience teaches us so much about the world, and that people should be able to claim that expertise, step forward, tell their stories, and share what they have learned. Because I see this as such a deeply human proposition, it has honestly shaken me to see AI in book publishing coming into authors’ projects in places I never anticipated.
This is the second part of my four-part series on AI and nonfiction writing. In the first part, I talked about why I do not recommend using generative AI to write your book and about the framework I keep coming back to: writing is a series of choices. In this episode, I want to move through the actual publishing process and show you the places where I have watched AI creep into real authors’ projects, from the first draft all the way through editing, cover design, and marketing.
What really prompted this series was realizing that the people using AI are often not people who are trying to game or cheat the system. In many cases, these are authors whose values I share and whose books are deeply important to them. The reason they turn to LLMs is the same thing I have talked about on this podcast for years: insecurity and fear.
Book writing pushes us out of the places where we feel secure, predictable, and safe. That is especially true when we are writing books that draw on our lived experience, where the whole process can feel like you are naked in Times Square. I understand why an LLM becomes appealing in that situation. If Claude can “clean up” your writing, maybe you do not have to sit with the discomfort of not being as good at this yet as you want to be.
Unfortunately, what I am seeing over and over is that pulling in an LLM to finish or supposedly improve the manuscript is never a good choice. I want to give you some real examples, not because I am interested in shaming anyone, but because I think seeing how this happens in practice can help you recognize it in your own process and make more grounded decisions.
Why AI Becomes So Tempting During the Publishing Process
Before I get into the examples, I want to explain one thing about the Page & Podium process because it matters here. Our publishing work is intentionally collaborative. First-time authors not only need industry professionals to walk them through this process; I think they deserve to understand why decisions are being made, what the parameters are, and how those decisions affect the finished book.
That is a big part of how we approach publishing at Page & Podium. I never want an author to feel like the book has been taken out of their hands. Our job is to give them the information and support they need to make informed decisions in a process that is, frankly, vulnerable and sometimes scary.
The examples I am going to share come out of that workflow. Different presses will handle things differently, but these are situations I have actually seen. I also want to be clear that I do not blame authors for being tempted by these tools, particularly when we are constantly being told that we will be “left behind” if we do not get on board with AI.
Friends, I do not believe that is true.
In fact, when it comes to my own writing, I am perfectly comfortable being left behind by AI. I am a better writer than Claude. And I think the examples below make a pretty strong case that LLMs are not going to make your nonfiction book better either.
AI in Book Publishing Stage #1: Writing the Manuscript
The first place I want to start is the writing itself.
Most of the authors I talk to are not planning to use AI to write their entire manuscript. Usually, they imagine using it for bits and pieces. But I did work with one author who appeared to be using an LLM to generate essentially the whole manuscript.
She would ask Claude or ChatGPT to write a chapter and then copy and paste the output into Microsoft Word. I did not know that was happening at first, but it became obvious very quickly because I was seeing the same choices over and over again.
If you listened to the first episode in this series, you know that this repetition is one of the signatures of LLM writing. A large language model is designed to make highly probable choices, so when it finds a pattern that works, it tends to keep returning to that pattern. Human writers generally make a much wider range of choices.
In this manuscript, I would see the same sentence appearing more than once, but the bigger problem was the repeated formula. Page after page used nearly the same structure: setup, bold declaration after a colon, then a list of synonyms or near-synonyms. The information itself also became repetitive and, at times, contradictory. Eventually, the manuscript was so repetitive at the sentence, paragraph, and page level that there really was no way to edit it.
It needed to be rewritten.
This is one of the issues I have written about before in AI Is Killing Your Writing. When the underlying choices are repetitive, line editing cannot fix the problem because the problem is baked into the structure of the prose itself.
Can using AI make it harder to write without it?
That experience also raised a much larger concern for me. I suspect there is an amount of LLM use that eventually makes it harder to produce your own writing.
Someone on LinkedIn once shared an analogy with me that I found incredibly helpful. Think about GPS. Before GPS was everywhere, a lot of us simply knew how to get around our cities. We knew different routes, how the streets intersected, and where things were because we had to.
I do not know my city that way anymore. I use GPS for almost everything.
Why would we assume that GPS can reduce our ability to navigate, but an LLM cannot reduce our ability to write? The mechanism seems pretty similar to me. Writing is a series of decisions, and if we repeatedly outsource those decisions, we should at least consider the possibility that we are also reducing our practice making them.
That is why I am so cautious about using AI during drafting. You build your writing ability by actually writing, just as you build your sense of voice by making your own sentence-level decisions. I have written more about that in Sentence-Level Editing: 3 Simple Ways to Strengthen Your Writing and in What Makes a Memoir Good?.
AI in Book Publishing Stage #2: Editing Your Manuscript
The second stage is editing, and I want to give you two very different examples because AI can come into this part of the process in ways that seem much smaller and more harmless.
The first is an author I am working with now whose manuscript is genuinely excellent. She submitted it for a Manuscript Review, and as I read, I could tell there were sections that felt very human and other sections that felt dramatically different. She later confirmed that some of those passages had been generated by AI.
When we met to discuss the review and make a revision plan, she told me she had gone back and reread the passages I flagged. Her reaction was essentially, “You’re right. They do not sound like me.”
They did not.
Now we are working to de-AI-ify those sections, which is absolutely possible. But the interesting question to me is why they had sounded good enough when she first put them into the manuscript.
AI solves the blinking-cursor problem
A client said something to me recently that really clicked: AI gets rid of the blinking-cursor problem.
That is true. The blinking cursor is a real psychological problem. You are staring at a blank page, trying to figure out what you want to say, while simultaneously talking yourself out of whether you are capable of saying it.
Then you ask an LLM to write it, and suddenly the page is full.
Those words may not be especially good, but they are also probably not as bad as you feared they would be. The output appears almost instantly, which makes the whole thing feel a little magical. You drop the paragraph into the manuscript and move on.
Weeks or months later, when you come back to the passage, the magic has worn off. Suddenly, you are looking at the actual writing rather than the relief of no longer staring at a blank page, and the passage does not seem nearly as impressive.
This is where I think writers need to be very careful. The moment when you cannot figure out a paragraph is often the exact point where learning is happening. If you bail out every time something becomes uncomfortable, you never build the ability to work your way through that kind of problem yourself.
I have talked elsewhere about how professional writers work through writer’s block and why I do not think the answer is simply to “push through” in a punishing way. The point is not to suffer for the sake of suffering. The point is to stay engaged long enough to figure out how you solve the problem. Our piece on why we should retire the phrase “push through” gets into that distinction in more detail.
Writing a book, especially a memoir or another book rooted in lived experience, requires building your capacity for discomfort. If AI removes the discomfort every time it appears, it also removes the practice you would have gotten by finding your own way through it.
Do Not Use AI to Explain Why You Disagree With Your Editor
The second editing example is a very different kind of problem.
We had ghostwritten a memoir for a client whose story included significant trauma, including sexual trauma, domestic abuse, and intimate partner violence. It was a very powerful book, and the author had been extremely happy with the manuscript throughout the process.
When our ghostwriter reached the final chapter, she made an interesting structural choice. The memoir had been written in first person, but the final chapter shifted into second person so that the author could speak directly to the reader.
I really liked the choice. Second person is not something we invented at Page & Podium; it is a literary device that has been used very effectively in other books, and in this particular manuscript I thought it made the connection between the author’s experience and her message to the reader especially direct.
The author did not like it, and that was completely okay.
When we do collaborative ghostwriting, the author does not have to accept every choice the writer makes. Usually, if something feels wrong, the author brings it back to the ghostwriter and tries to explain what they dislike about it. That can be difficult when the author is not a writer and does not have the terminology for what is bothering them, but that struggle is actually useful to us.
Sometimes “I don’t like this” is only the surface reaction. The conversation helps us figure out what is underneath it so that we can solve the real problem.
In this case, though, the author put the chapter into an LLM and told it that she did not like the use of you. She asked AI to explain why she did not like it.
The system responded with a very clean, very confident answer saying that second person should never be used in memoir, that it was inconsistent with the genre, and that it was bad writing.
Now we had two problems. We still did not know what the author actually disliked about the chapter, and we also had to reassure her that the ghostwriter had not made a careless or uneducated choice.
That is why I do not recommend using AI as a referee between you and your editor, ghostwriter, or coach.
If you receive editorial feedback you dislike, talk about it. There may be another edit that solves the same reader problem. You may learn something about why the original suggestion was made. You may still decide you do not want the change.
All of those possibilities are available when you actually have the conversation.
I have written before about when to get feedback on your memoir, and one of the central ideas there is that good feedback should deepen your understanding of the manuscript. It should not simply hand down a verdict.
AI in Book Publishing Stage #3: Cover Design
The third stage is design, especially cover design.
Book covers are incredibly personal for authors. They are quite literally the face of your book, so people become understandably emotionally invested in them.
That emotional investment also makes cover design a perfect place for AI’s certainty to meet the author’s insecurity.
One of our partnership publishing clients had moved all the way through an excellent manuscript and reached the cover-design stage. In our process, the designer creates a series of concepts, which are not final covers but are developed enough to give us something concrete to work from. The author chooses the concept they like best, and then we make adjustments together.
This particular author initially loved the concepts. He chose one and told us how impressed he was with the designer’s work.
Then, as we moved through the revision process, requests started appearing that had not really been part of the original conversation. Eventually we discovered that he had put the cover concept into AI and started coaching the tool to make changes.
The cover he sent back was completely different.
His book was broadly a self-help book aimed at young adults. The AI-generated version looked much more like techno-dystopian fiction. It might have been visually dramatic, but it did not fit the genre or communicate what the book actually was.
This is where comparable titles matter.
At the beginning of our cover process, we show authors covers from books that sit in a similar area of the market. A comparable title is simply another book that is similar enough to yours that we can use it to understand reader expectations. We are not trying to copy those covers, but we do need your book to exist in the same general visual world so that the reader can understand what they are looking at.
That is one of the reasons professional publishing involves more than simply creating a cover the author likes. The cover needs to work for the reader and for the marketplace too.
When you put a cover into AI and start telling it what you want, the system is very happy to help you visualize the idea already in your head. What it is not going to do is remind you that your cover is not only for you.
In this case, we were able to go back to the original professional concept, make some tweaks, and end up with a cover the author loved even more. The AI detour did not ruin the project, but it did not help us get where we needed to go either.
AI in Book Publishing Stage #4: Marketing Your Book
The final place I want to talk about is marketing.
Page & Podium works with a lot of experts and people who have a message that is deeply important to them, so content marketing is a big part of the strategy for many of our authors. That might mean a podcast, YouTube channel, newsletter, social media, speaking, or some combination of those things.
Content marketing works because you are creating material around ideas you genuinely care about. Readers encounter those ideas, connect with them, and some of them eventually want the deeper version in your book. It can also be incredibly useful for the author because the act of creating content keeps the ideas alive and keeps you developing them.
Unfortunately, this is another place where I am seeing AI interfere with the process.
The obvious problem is outsourcing your entire voice to an LLM. Anyone who has spent time on social media lately has probably seen plenty of that.
The bigger issue, in my opinion, is outsourcing your thinking.
My friend Jess Sato works with thought leaders to make sure their big idea, mission, and vision stay central to their businesses. Like me, Jess has spent years creating content and actually enjoys doing it.
Recently, she talked on her Profit and Purpose Podcast about what she calls an “AI cleanse.” She had been using AI, as I hear people say all the time, “just to generate ideas” and “just to organize her thoughts.”
Over time, though, she noticed that she was having a harder and harder time coming up with topics herself. Eventually, she felt like she did not have ideas unless she went to AI first, and the ideas AI was producing were not especially exciting to her either.
When she stepped away from it, she found that her own ideas and enthusiasm for her topic began to return.
That example struck me because I hear the word just all the time when people describe their AI use.
They are “just” using it to brainstorm or organize ideas.
But very often, the thing that comes after just is much more central to the creative process than we realize.
When you organize your own ideas, you find connections between them. One theme reminds you of another. You remember an example you had forgotten. A new question occurs to you because you are physically moving through the material yourself.
Those discoveries are part of the thinking.
When we outsource the organization because AI can do it faster, we also lose the opportunity to make those connections ourselves. That becomes especially important for writers who are trying to build expertise and authority through their work. We have talked about developing a book idea and about how a book can grow your business, and both depend on the author actually developing their thinking rather than simply producing content.
If you use AI every time you need a topic, eventually you may find that generating topics without it becomes harder.
The Problem With “I’m Just Using AI for…”
This is the pattern I want you to notice across all four stages.
During drafting, somebody is “just” asking AI to get them past one difficult paragraph. In editing, somebody is “just” using it to explain feedback they do not like. Once you get to design, somebody is “just” trying to show the designer what they have in mind. During marketing, somebody is “just” asking for a few ideas.
In each case, the supposedly small task involves choices that matter.
That is one reason I am so insistent about writers doing their own development and structural work. When you outline a nonfiction book, the finished outline is only part of the benefit. The process of figuring out which ideas belong together, how the reader should move through the argument, and what each chapter needs to accomplish is itself important.
The same thing is true of memoir. An effective memoir outline helps you identify the topic, message, turning points, and progression of the story. If a computer gives you the finished outline instantly, you may have the document without having developed the understanding behind it.
Revision works the same way. A top-down revision process requires you to step back, look at the manuscript as a whole, and understand what it is doing before you start tweaking sentences.
The thinking is part of the work.
What If AI Is Already Deep in Your Book?
If you are reading this and recognizing yourself in one of these examples, I do not want you to feel ashamed. I understand why authors turn to these tools, particularly when the cultural message is that AI is inevitable and that refusing to use it means falling behind.
You can back out of it.
I am already working with authors who are doing exactly that.
Sometimes that means rewriting AI-generated passages so that they actually sound like the author again. In another manuscript, it may mean rebuilding the structure so the writer understands why every chapter is there. Someone else might need to return to editorial feedback and have the conversation they skipped the first time.
The larger goal is to rebuild your confidence in your own creative decision-making.
If you have a manuscript and no longer know what is working or where AI may have taken the project off track, our Manuscript Review is designed to give you that 10,000-foot assessment and a step-by-step plan for revision. If you are earlier in the process, the Memoir Method Checklist lays out the full path from idea to finished book so that you can understand what kind of work belongs at each stage.
There really is no shortcut around making the decisions that create the book. But the good news is that the more you make those decisions, the better you get at making them.
My Recommendation on AI in Book Publishing
After seeing AI in book publishing appear in writing, editing, design, and marketing, I keep coming back to the same thing: AI is most tempting when its certainty meets our insecurity.
When you do not know how to write the paragraph, it can write one immediately. If you are unsure whether your editor is right, it can give you a confident answer. When you cannot articulate the cover in your head, it can generate an image. If you have no idea what to post this week, it can produce twenty topics in seconds.
The problem is that becoming an author requires learning how to make decisions even when you are not completely certain.
My professional recommendation remains the same as it was in the first part of this series: do not use generative AI to do the creative work of writing your book. I would also extend that caution beyond the drafting stage. Using AI to evaluate your editor, redesign your cover, or organize your marketing ideas may not look like “AI writing,” but those activities still include decisions that are part of the creative and intellectual work.
Before you outsource something, ask what decision the tool is actually making for you. Consider whether working through that decision yourself, or collaboratively with another human, is how you develop the skill and confidence you need.
The goal is not simply to end up with a book.
I want you to emerge from the process more confident in your own ideas, your creativity, and your ability to write the next one too. That is a much better outcome than getting to the finish line faster.
