As a three-time published author and ghostwriter for clients, I thought I was really onto something. I even changed my LinkedIn headline to reflect this breakthrough: “Award-Winning Author + AI = 60-Day Book.”
Then I sold my services to two clients. One wanted to tell the story of a historic advocacy campaign her organization ran. The second is a very in-demand therapist who is working to share his unique model far and wide.
Let’s just say that AI helped me turn around these books, but not at all in the way I expected, and not in sixty days either.
These projects taught me in a BIG way that professional writers are here to stay. I also learned that before you approach a lengthy writing project with generative AI, you should have a clear strategy for how you’ll use AI that you can share with your clients. This strategy needs to be based on the technology’s strengths and weaknesses rather than on your hopes and dreams.
With that in mind, here are some things I discovered through working on these two books:
1. AI didn’t make me faster, it made me better
The real breakthrough wasn’t speed—it was research capability. Perplexity.ai surfaced obscure documents that would never show up in Google searches, such as court filings that were buried in legal databases, and academic papers that were hidden in institutional repositories.
But Perplexity was only about 85% accurate about what those sources actually contained. It would confidently state, “This document shows XYZ,” but then when I clicked into the document to pinpoint the data, I realized the AI had misrepresented what the document said—or the “source” didn’t contain any material close to what Perplexity had promised at all. So while this chatbot pointed me in a direction that was often helpful, I couldn’t take its word for anything.
That 15% margin of error? It’s dangerous territory when your client’s credibility is on the line. For this reason, I vowed to never take AI’s word for anything. My rule? When it comes to research, never trust AI. If I want to include a fact, I need to track it down in the primary source document with my own human eyes.
2. AI has verbal diarrhea
On the plus side, generative AI chatbots never get tired. On the minus side, they spew out endless text without weighing its value and without paring it down to the essential nuggets of information that will resonate with the target readers.
I learned this the day I decided to wage a robot fight.
Up to this point, I hadn’t been using AI to generate extensive copy. Then one day, I got frustrated with how a particular chapter was coming out. As an experiment, I decided to let AI write it almost completely without me.
I created two roles in ChatGPT—an expert book editor and a copywriter—and had them argue back and forth for four hours. To kick things off, I fed AI some chapter notes my client had supplied.
Me: “You are the expert book editor. Use these notes to create a chapter outline.”
Book editor creates the outline.
Me; “Switch roles. You are the copywriter. Implement the editor’s notes. Write the first section.”
Copywriter implements notes.
Me: “Switch roles. You are the book editor. Rate the copywriter’s work 1-10 and make suggestions.”
Book editor rates the work and makes suggestions.
Me: “Switch roles. You are the copywriter. Implement the editor’s suggestions.”
Back and forth, back and forth.
I was watching two very polite robots have a never-ending conversation. After four hours, I did have a chapter draft. But it was completely unwieldy and still needed total restructuring by a professional human writer who actually understands how ideas flow and connect.
3. AI is a brilliant brainstorming partner with terrible ethics
When I started working on these books, I was hoping to have the AI comb through pre-existing materials my clients had written such as articles, speeches, and books. I wanted to be able to say, “Find me everything my client has written about X in these seventy sources.”
I tried a variety of chatbots from ChatGPT to Claude. While the chatbots were great at weaving together information from disparate sources, they were even better at interspersing real facts with fabricated ones.
Over time, my threatening prompts evolved like this:
Prompt: “If there is missing information or you need clarification, please ask.”
Prompt: “It is wrong to make up missing information. Do not do it.”
Prompt: “If you make up any facts, I will fine you $2,000.”
None of this worked, so as a result, I needed to spend extra time fact-checking every word and making sure that every sentence truly did come from my client’s previous work.
If you are a subject matter expert, your knowledge base deserves to be mined for gold! Your previous publications, if you have them, probably hold incredible insights.
Someone needs to mine that gold and share it with your target audience in a way that resonates. And that someone, I’m quite confident, is human.
You really need a lion tamer: here’s why
If you’re hiring a ghostwriter for your book, look for a lion tamer—an expert with enough confidence and experience in their writing skills to tame AI into submission. It’s easy to see how someone without the deep knowledge of how to shape a manuscript could be led by AI rather than whipping it into submission.
Find a writer who knows when to trust AI’s output and when to fact-check everything back to the original source. Look for someone who knows how to wrangle the technology and can surface AI’s creativity, while also maintaining professional judgment about how to shape content so that it’s most poignant.
Don’t risk your thought leadership on unvetted AI output. The stakes are too high, and the technology isn’t there yet.
In the end, the two book projects took me about four months each, which was double the time I had expected. This happened for a variety of reasons, not all of which had to do with AI.
But now? I am rock-solid confident in the accuracy of the information and the storytelling quality in each book; each client’s trademark ideas and style reverberate on every page.
I’ve also taken down my old LinkedIn headline. I’m currently searching for the right words to express that I’ve jumped through the ring of fire and learned to tame the wild AI beast. Would love your suggestions!
Need help with your book or fact-checking your AI work? DM me.
Shana Burg is a Level 7 ghostwriter and the author of “Poof! The Disappearance of a Writer in the Age of AI.” She holds a Master’s in Public Policy from Harvard.