Did AI Write My AI Book?

I haven’t been a published author for very long (the book just came out today), but the number one question I’ve gotten so far about the book is, “Did AI help you write it?”

AI did not write The Non-Obvious Guide to Using AI for Marketing. For better or worse, it’s my work.?

I actually used AI to write two other books and even had them on Amazon for a bit to understand the process end-to-end. I spent roughly one hour collectively creating and publishing both. They’re no longer for sale, as I got what I needed out of the experiment, and I don’t need to contribute to the AI slop problem. It was a useful exercise largely because now I’m one of relatively few people in the world who has published a book under my name written by me and written by AI, and I can tell you about the differences so you don’t have to do either.

It’s hard to claim Using AI for Marketing its mine alone. There are editors, guest interviewees, examples from a bunch of brands, and others who contributed – directly and indirectly. Plus, I run a community for marketers using AI, so I’m learning from members all the time – directly and indirectly.

Could I have written this book without the use of generative AI at all? Absolutely.

Would I recommend avoiding AI entirely when writing a book today? Absolutely not.

Writing a book without Google would have been a slog, as that would require library science skills that are too rusty (if I ever had them). Doing so without Grammarly would have provided my editors with undue aggravation; they had enough as is (sorry for some of those oversights, Kameron).?

But without generative AI? I’ve written well over 1 million words in work that I’ve published in various outlets, all before the age of generative AI, so writing tens of thousands of words that way wouldn’t be too hard of a challenge.

Generative AI did enhance the process for me. It also serves as a primer for some of the AI applications that I discussed in the book. I have a brief section about this on the book, but almost everything above and much of what’s below is different from the published version. If you want a more concise version, buy the book.

Here’s how I worked with AI:

1. Brainstorming the book proposal. I used AI engines to come up with the initial proposal draft. I had some writing and presentations to upload so the AI (mostly ChatGPT and Claude) could get a sense of what kinds of topics I’ve discussed. I also shared collateral about AI Marketers Guild so AI engines knew more about my connection to the subject.? One version of a proposal I submitted had specific requirements, so I used AI to make some sections longer, shorter, or better tailored to the questions. As I edited the proposal, it became more useful. For instance, I used AI to take the finished proposal and turn it into a brief synopsis of the book for marketing purposes; this would later be reworked once the book was written. ?
2. Filling missing gaps. After coming up with certain lists, such as the audit for overcoming organizational resistance to AI, I’d ask an AI engine for other ideas to see what I might have missed. I would specifically tell it to give me a list and not detailed explanations. One rule of mine writing the book was that AI could help me brainstorm, but all the writing had to be my own.? That also meant rejecting most of AI engines’ ideas. Some of those ideas didn’t make sense, but sometimes there was the subjective rationale where it didn’t feel like something I’d write about. The book had to be “me” as much as possible, even if books always benefit from researching the work of others, improvements by editors, and all kinds of inspiration. ?
3. Adding research and examples. Often, I’d use AI to look up data and concrete examples. The AI engine Perplexity was most useful here because of its deserved reputation for meticulously documenting its sources. In the chapter on overcoming objections, for instance, I wanted to know which research studies were conducted showing how consumers feel about marketers using AI. I could have done this with Google’s search engine in the past, but AI engines typically organize answers better and in a way that directly answers your question. They often know the gist of what you’re talking about even when you don’t articulate the perfect wording.? In all such cases, just as one shouldn’t use search engine results listings or Wikipedia as a source, I went to the original sources (and documented those for my editors and me) to review every fact cited by AI. I also constantly would copy in sentences I wrote and ask AI engines, “Is this accurate?” Again, it was my call how to make sense of its judgment and whether to make changes. I can tell you that the book does not have hallucinations. It might have bad takes or ill-informed opinions – and I hope there’s at least something you disagree with. If a factual error is in the book though, it’s mine.? ?
4. Organizing the book: I used private, custom workspaces (ChatGPT’s Custom GPTs and Google’s NotebookLM) to keep track of my writing. This was useful when encountering topics and figuring out how to work them into the book. For instance, I read a great article about the role of the Chief AI Officer (CAIO), and I was wondering whether I had written about this yet (no judging, please; I was working on multiple chapters at once over several months). It turns out that I hadn’t yet included it. But then I could have a discussion with AI about where this section could go, while debating pros and cons.? ?
5. Adapting content for online resources: For many of the same reasons that AI doesn’t come up with original ideas (all the output is determined by algorithms), AI excels at repurposing your own content. It is great for taking my writing and turning it into other formats like checklists and quizzes. I do this all the time for my marketing consulting. For instance, I might have a long conversation over Zoom about a document I’m working on with a client or colleague, and then I’ll take the transcript of that call and have an AI engine draft the text for a single slide based on everything we discussed on the call. AI tends to be most useful when it’s not ‘thinking’ but organizing. AI helped format my writing to work as other kinds of resources, but it didn’t write those resources itself. ?
6. Marketing the book: I wish AI could market the book for me on its own. I could give it the budget and some parameters, and then let it rip, while telling it to optimize around certain metrics. It’s not quite that easy. But I’m keeping AI busy as my focus shifts from producing the book to marketing it. Now that I’m adapting content for different formats – resource guides, interviews, presentations – AI is my assistant. And it is also a second brain as I resurface which examples and quotes I used in the book. A favorite example so far was when I started working on my master ‘stump speech’ deck about the book that I can adapt, I asked NotebookLM for all the funniest examples in my book. It brought up one, for instance, about NYC botching its AI rollout, and few things are more universally entertaining than dumping on Mayor Adams. Google gets me.

It is now impossible to write a book without AI, short of some Thoreau approach with a typewriter or handwritten tome that you compose in the woods. If you have the fortitude to do so, I’d recommend picking a topic other than AI to write about.

Should you accept that you will be AI to assist the writing of it, you can at least avoid having AI write the work for you. You’re probably a better writer anyway – and undoubtedly a more interesting one.