What I Actually Tell Clients About AI

People ask me about AI all the time. On sales calls, at conferences, in the middle of working on their book when something they read scared them. The questions repeat. The answers I give have settled, across years of these conversations, into something I’d write down if anyone asked me to. Here are six of them, and exactly what I say.

Lisa Hamilton-Smith, a life coach who now teaches AI to solopreneurs, does the same with her coaching clients. She tells every one of them their material is going into AI, and she said on Leaders and Their Stories that 99.9% of them have no problem with it once they see what it does. Telling them first is the whole trick.

“Can I just have ChatGPT write my book?”

You can. People do. I’m going to tell you what happens when they do, because I’ve watched it happen more than once.

A client of mine, before I started working with him, drafted his whole book in ChatGPT 3.5. He was proud of it. He sent me the chapters expecting me to be impressed. They were terrible. They weren’t offensively terrible. They were hollow, the way every AI-written long-form work is hollow. Clean grammar, polished sentences, every paragraph sitting politely next to the last one, and nobody home.

I didn’t tell him the chapters were bad.

I ignored them, ran a real interview series with him, and started writing the book from his actual life. He kept sending me more ChatGPT chapters, trying to help. He liked the AI versions. The whole project cost extra back-and-forth time because he had to be talked off the AI drafts every few weeks, and the time was billable.

This is the version most people who ask the question don’t picture. They picture writing the book themselves, with AI as the typist, finishing in a month. The machine produces a book that reads dead, the writer can’t see why, and they ship a book that damages their reputation instead of building it. The full case is in It’s Not the AI, It’s the Crap.

No. You don’t want ChatGPT to write your book. You want it to help you organize the research and outline the chapters. The writing has to come from you, or from a human you trust.

“Should I use AI for the research part?”

Yes, with one rule: verify every specific claim before it goes in the book.

AI is useful for research synthesis. You can have it read ten papers on a topic and summarize what they all say, find the consensus, name the disagreements, point to the gaps. You can have it help you compile what’s been written about your subject across dozens of sources. You can have it draft a working outline for a chapter based on a transcript of you talking through what you know about a topic.

What you can’t do is trust the facts it produces. AI invents citations that don’t exist. It misattributes quotes. It gets dates wrong. It states statistics with full confidence that came from nowhere. Every name, date, number, quote, or citation in your AI-generated research has to be verified against an independent source before it ends up in the book. If you can’t verify it, you cut it.

This is slower than just trusting the AI. It’s also the only thing that prevents you from publishing fabricated facts under your own name. I covered the mechanics in AI Hallucination: A Survival Guide for People Who Publish Under Their Own Name.

“My grandfather/mother/partner is in their seventies and wants to write a memoir. Is AI useful for them?”

Yes, and they’re in a better position to use AI well than most people half their age.

I work with a client in his seventies on a memoir. He uses AI in three specific ways.

He uses it to find missing bits in his own draft, the chapters where the pacing is off or where he hasn’t said enough about something. When he can’t remember whether a story belongs in chapter three or chapter seven, he uses it to focus himself.

He uses it to generate notes for me about what he wants me to work on next. When we talk on Tuesday he has a clear agenda, instead of a vague feeling about what isn’t right.

He doesn’t let it write any of his book. He’s adamant about that. Every word in the manuscript is his, because the book is his life and his voice and the machine has no business inventing either.

That’s the model. The machine handles the logistics of writing.

The human handles the writing. For an older writer working on a memoir, AI is useful as an organizing and focusing tool. It’s useless as a writer. The 71-year-old client knows the difference. He’s going to publish a better book than people half his age who don’t. The longer version of his story is in The 71-Year-Old Memoirist Who Uses AI Better Than You Do.

“Will my readers know if I used AI?”

If you used it to organize your research and outline your chapters, no, and they shouldn’t care. That’s logistics. Every working writer uses tools for logistics, and AI is the current best tool for some of them.

If you used it to write the prose, yes, and they’ll care.

Readers can hear it. They can’t always name what they’re hearing, but they can hear it. The technically clean sentences that have nothing in them. The chapter structure that sits exactly where you’d expect. The absence of the risky detail that only someone who lived the work would know to include. The reader senses something hollow and stops trusting the book. Sometimes they finish anyway. They don’t recommend it. The book doesn’t do what it was supposed to do.

Disclosure wouldn’t have fixed anything. The reader rejected the book because the writing was hollow, and the machine just happened to be the explanation they found when they went looking.

The line I draw with my own clients is the same one I draw for myself. AI is fine for organizing interviews, building summaries, drafting outlines, handling communications. I go all the way up to the writing and stop there. I’m not going to thread 20 interviews into an outline by hand, and I don’t even know how I did that in the past.

The grammar checkers are fine, but I never accept their replacement text, because that’s where the machine voice creeps in. The book itself, every word, is human. I hold that line because it’s the only thing that produces a book worth reading.

“My ghostwriter says they don’t use AI. Should I believe them?”

It depends on what they mean by “use.”

If they mean they never touch AI for any part of the work, including organizing interviews, summarizing research, or drafting their own communications, they’re either telling the truth and working slowly, or they’re lying. Both are possible. Neither is necessarily a problem.

If they mean a human writes the prose of your book, that’s the answer you want, and that’s the answer that matters. Ask them this question directly. “Will the words in my book be written by you, or will any of them be generated by AI?” A good ghostwriter will give you a direct yes-the-prose-is-human, with whatever caveat is honest about their workflow (most ghostwriters use AI for logistics; very few use it for prose).

A ghostwriter who uses AI to organize interviews is fine by me. Worry about the one who can’t tell you exactly where they draw the line, or who gets defensive when you ask. The line is what you’re hiring, and a writer who won’t say where it sits is hiding something about how your book gets made.

“I’m too old / too busy / too far behind to learn AI. Is that okay?”

No. My reason is competitive, and it has nothing to do with morals.

It’s not okay because the people in your industry who are willing to learn are going to outrun you, and they aren’t waiting for you to catch up. The longer you wait, the further behind you get. The further behind you get, the harder it is to start. By the time you finally start, you’ll be starting against people who are now five years sharper than you on tools you’ll be learning from scratch.

I’m not telling you to become an AI expert. I’m not telling you to spend your weekends reading research papers. I’m telling you to pick one boring repetitive part of your work this week and find an AI tool that handles it. Spend an hour learning. Spend the next two weeks using it. Notice what time you got back. Use the time to get sharper at the part of your work that only you can do.

Repeat that next month with the next boring part of your work. In six months you’ll be augmented. The full version of this argument is in Retrain or Be Replaced. There’s No Third Option.

What it adds up to

From the podcast

Diana Lee has built roughly eighty micro tools for exactly this kind of use, and frames the value the same way on Leaders and Their Stories: If you use it in a way to amplify you in the way that you’re using it, I think it’s the self-awareness piece that’s necessary to just be open to hearing how things can be done better. The people who get the most out of these tools are the ones willing to hear that something isn’t working.

Use AI for the logistics and keep it away from the writing. Verify every specific claim before you publish it. Learn one thing this week instead of promising yourself everything by next year. Pick a ghostwriter who can tell you exactly where the machine stops. The authors who ignore this put their names on hollow books, and they pay for it with their reputations.

That’s what I tell every client who asks. The questions are different. The answers settle into the same shape. The Birth of the Augmented Human covers the broader version of what this looks like across every profession, but for writers and authors and people working on a book, the rules above are the working version.

From the author

The Ghostwriting Advantage
Clients bring me their AI questions before they bring me their stories. The Ghostwriting Advantage is the book I wrote on how to get your book written for you.

Frequently Asked Questions