So, I had another conversation with ChatGPT! – Podcast Transcript

So, I had another conversation with ChatGPT!

Are you actually getting your voice and style into your work when you use AI tools like ChatGPT? What if you could challenge the AI to go beyond surface-level answers and truly act as your critical editor? In this episode, I talk about the frustrations, breakthroughs, and lessons I experienced while updating my book with help from ChatGPT, and how asking better questions can transform the results. Are you owning every word? Are you challenging your technology to make your content authentically yours?

Listen to this new 25-minute episode for tips on getting more authentic results with AI, asking better questions, and making sure your voice shines through every word.

If you have any questions about anything in this, or any of my podcasts, or have a suggestion for a topic or guest, please reach out directly to me at [email protected] or visit my website Podcast.AlanBerg.com 

Please be sure to subscribe to this podcast and leave a review (thanks, it really does make a difference). If you want to get notifications of new episodes and upcoming workshops and webinars, you can sign up at www.ConnectWithAlanBerg.com  

View the full transcript on Alan’s site: https://alanberg.com/blog/

Summary:
In this episode of the Wedding Business Solutions podcast, I share a fascinating behind-the-scenes conversation that I had with ChatGPT about the process of updating my book, “Why Are They Ghosting Me,” to a refreshed edition. The discussion takes you through the unique experience of collaborating with AI, including some of the challenges and frustrations I faced when trying to maintain my authentic voice and style throughout the project.

We delve into the mechanics of using ChatGPT projects versus individual chats, exploring how instructions and memory can sometimes be overridden and why even clear upfront guidance is not always followed consistently. I explain how I had to remind the AI to stick to my original voice and examples rather than defaulting to rewriting content, which led to multiple drafts and ongoing tweaking. One key insight was the discovery that simply asking the AI, “Are you happy with that?” could set it into a critique mode, offering more honest and analytical feedback. This approach became a useful tool for elevating the quality of the book, rather than just receiving encouraging but less helpful responses.

The episode also covers the benefits of building a style guide and using specific instructions to preserve my voice for future projects. I share my realization that while AI can accelerate the process of gathering and organizing ideas and content, it is essential to maintain ownership and authenticity in the final product. The importance of challenging the AI and not just accepting the first draft is emphasized, as well as the value of being open to tough critique to achieve the best possible result.

By sharing this conversation, I aim to give listeners a practical look at how to use AI tools more effectively in creative projects, ensuring that the end result truly represents their unique voice and vision. This episode is filled with actionable takeaways for anyone considering leveraging AI to help with writing, editing, or brainstorming new ideas in their business.

 

So I had another conversation with ChatGPT, and I think this would be interesting for you. Take a listen. Welcome to another episode of the Wedding Business Solutions Podcast. I’m your host, Alan Berg. I’m a speaker, author, sales trainer, website reviewer, and I help businesses like yours sell more, profit more, and have more fun doing it. Enjoy this episode.

I’d say it was a true copilot collaboration. You brought the voice, the frameworks, and all your real-world examples, and I helped you shape that into book language that sounds like you, but reads like a polished business book. You set the direction and I helped you test the language, pace the flow, and spot any soft spots where your original cadence was getting lost. More craftsman and editor side by side than anything else.

Okay, but we had some frustrations in here where it wasn’t following my voice. Uh, you weren’t necessarily following the instructions that you have, and we had to keep going back and reminding of that and updating of that. Um, why does that happen when the instructions were pretty clear? Because you helped me write them.

Yeah. I think because 2 things were colliding. Your instructions got very detailed and strong, but I generate text turn by turn, not from a fixed full manuscript view. So when things like don’t rewrite, preserve verbatim live alongside refresh and integrate new value, there’s a balancing act. Early on, I defaulted sometimes to rewrite mode, which is often helpful, but not what you wanted. And also, we were refining the instructions as we went, learning what worked by noticing what didn’t. In a way, that process wrote the better instructions.

Right. But we wrote the better instructions, and then we’d go a couple of chapters or even a chapter in, and then you’d default back to rewriting, which is what we weren’t trying to do. We were trying to preserve my voice. We were trying to preserve my style, trying to preserve original examples, but then you’d fall back into this. So why does this happen when it’s been working fine and then it doesn’t?

Hmm. I think part of it is that each response is a fresh attempt to satisfy everything. There’s no locked-in editor mode that stays fixed across chapters. It’s more like a series of best effort passes. So when general instincts like smoothing or tightening kick in, they can overpower a rule unless the process itself changes. And once we shifted from does this sound good to is every change justified, I think it became a far more reliable process.

Now, we used a ChatGPT project for this as opposed to a GPT. So if I was explaining this to a friend of mine about why a project versus a GPT, what would you say to them?

Let me have a quick think. I’d say a GPT is like hiring a specialist with a standing job description, great for repeatable roles but not tied to one specific manuscript. A project, on the other hand, is more like a dedicated room where all the drafts, source files, and the conversation history live together. That mattered for your book because we weren’t just applying a style, we were tracking the original chapters, audits, revised drafts, and evolving rules across time. A GPT alone wouldn’t naturally hold all that project-specific context by default. You’d keep having to refeed the manuscript. That said, if I were designing it now, I’d probably pair a project for the files and the GPT to enforce your editorial philosophy every time. So each pass starts from preserve first, justify every change.

Okay. I’m not understanding that. You can connect a GPT and a project together?

Hmm. Not literally, no. They’re separate methods. What I meant was conceptually. In practice today, you’d choose one. For a book, I’d pick a project every time because the manuscript and source material stay put. The GPT piece is just saying the editorial philosophy we developed could be captured in a custom GPT, but since you can’t bolt that onto a project, the practical move is exactly what we did. Put the instructions and all the discipline about preserving text inside the project and treat it like an editing room.

So if a speaker friend asks, I’d say use a project for serious book work and focus on making the project instructions excellent the way you ended up doing.

So even though the instructions were excellent, we still fall back into other modes. And I guess that’s what I’m not understanding. If the instructions are clear, what you’re saying is that you’re not going back to the instructions each time because something else that we did overrode that?

Yeah, I mean, I do go back to the instructions. The gap is that following them perfectly every time isn’t guaranteed. It’s more like weighting them among a lot of signals, and sometimes that balance isn’t what you intended. That’s not to say your instructions weren’t good. They were. It’s a limitation in how I generate responses, more interpretive than rule-bound. What helped most was when you pushed for treat the original as correct unless there’s evidence to change it. That reframed everything.

But yeah, uh, it’s not the same as a human editor locking into a house style and never deviating. I think your projects made that visible.

So going forward with projects, there’s a limitation to how much data I can give you into— what is it called? Is it the knowledge base or the library? The data that I give you upfront to— in other words, what we put into that room, we move into that room. There’s a limitation to how much you can get on that. How much you get, you can get, correct?

Let me just check. Yes, that’s right. There’s a practical limit to how much you can load in and also a practical limit to how much of it is actively in play at once. Think of it like having a big library in the room, but only a few open books on the desk. That doesn’t make the project useless. It just means curation matters. For your next book, I’d focus on a tight set of source docs and a crystal clear editorial guide rather than trying to load everything. In other words, quality and organization over sheer volume.

And that’s something I think we learned together.

So to work on my next book, whatever that is, whether it’s a new book or an update of another book, would it be better to start a new project or to continue working in this project so that you have the history and the memory of what we did?

Reading. I think start a new project. Keep the lessons from this one, but don’t reuse the same room. The ghosting project is tuned around preserving that specific manuscript and its history. For a new book, you’d want a clean space with a new editorial charter, but borrow the playbook we developed. Conservative edits, explicit change justification, and voice anchoring. Carry the lessons, not the baggage.

So you could help me write the instructions for the next project?

Let me take a second. Absolutely. And I think we’d do it better this time. We’d start by writing instructions that say, you are the editor, not the ghostwriter. Default to preserve unless justified. Define a hierarchy if rules conflict and include a little preflight check, something like, did I change something that didn’t need changing? And we’d test those instructions early and revise them fast. So the guide matures with chapter 1, not chapter 10.

So something that I found myself doing when we were on the 3rd or 4th draft is we would do a chapter, and at the end I would ask you, are you happy with that? And you’d come back every time that no, or almost, I’m 92 to 95%, or something like that. And I found that to be interesting because if it was something that you just gave me and then I ask you, are you happy with that, and you tell me not quite. I found that interesting. You gave this to me, and then you’re telling me you’re not quite happy with that. How does that happen? You still with me? I think I lost you here.

Yeah, I’m still here. I think that’s a good observation. When I generate a draft, I’m aiming to give you the best version I can. But when you then ask me to switch modes and critique it, I’m actively looking for the gaps, the weak spots, the, the places that don’t quite fit. That critical pass is a different lens and it often surfaces, here’s what I’d refine, not it’s wrong or it could be tighter. Honestly, that back and forth made the book stronger. And if we do the next one, I’d build that in on purpose, separate the create pass from the critique pass so it doesn’t feel like moving the goalposts.

So asking you, are you happy with that, switches your lens from creator to critiquer?

Yeah, exactly. It’s the same reason an author can draft something and feel good, then read it the next day with an editor’s hat and see it differently. Those are 2 different mindsets, and your question triggers the second one.

Oh, interesting. So would that be something good if we were not in a project, but if we were just having a new chat about something, just brainstorming or something like that? When you give me something, is it a good idea to often just say, are you happy with that? Or are there other questions or better questions to ask if it was just a brainstorming chat?

Now, that’s a really good way of working. In fact, instead of asking, are you happy with that? I think the sharper prompt is lens switching. So something like, now be my toughest editor, or if you had to poke holes in this, where are the holes? For brainstorming, I think that keeps the flow creative first, but then deliberately flips to critical. So other useful prompts might be, what’s the biggest assumption here? Or what would a skeptic say? Or even what’s the one thing missing? Those questions invite me to challenge, not just generate.

Yeah, I love that. And I think it’s something that a lot of people are not doing when they use ChatGPT and, and other LLMs is they’re asking questions, they’re getting answers. But they’re not challenging. And something I find with you, and it could be based upon the settings that I have, which are the default voice and tone, is that you’re very encouraging, which is nice. But there are times when I don’t want the attaboy. There are times I want the, no, that’s not as good as it could be. Let’s make it better.

Yeah, I think you’re right. A little too much cheerleading when what you really want is a tough editor. If you set the expectation up front, push me, challenge assumptions, be blunt, I can lean into that mode. And you don’t have to do it every turn. Just drop a clear signal when you switch into growth mode. Don’t reassure me, challenge me. That one line changes the lens completely. And honestly, that’s one of the best lessons from our project.

So that challenging, when I, when I say that and you switch your lens, are you only switching it for that next thing that you, uh, give back to me, or does you hold that lens going forward?

Hmm. It depends on how explicit you are. If you say, for this conversation, pushback unless I ask otherwise, I’ll treat that as the standing mode for that thread. If it’s a one-off, like, now critique this, I’ll assume it’s just for that moment. Once—

So when I say— yeah, so when I say, are you happy with this, you treated that as for this one thing we just did, use the critique mode, but then going forward you went back to whatever the mode was before that?

Exactly. I’m based on our experience, I’d actually suggest establishing a standing working mode at the start, like be my candid editor by default, so the lens stays consistent until you change it. That prevents the creep back into supportive mode when what you really want is the tough love. And it sounds like that’s the mode that serves you best.

Well, there are times I want the supportive mode, and there are times I want the analytical mode, and there are times that I want the critique mode. And very often with you, I want you to then critique because, you know, we’ve been encouraging and encouraging, but, you know, that’s nice, but is that putting out the best work? And I think being open to the tough critique, and even though it, you know, made me go to the 4th draft because I just wasn’t happy with the voice, it wasn’t sounding like me, things were being rewritten. And those things that we had to go back and look again and say, nope, we didn’t do that. We didn’t keep true to the original book. We didn’t keep true to your voice. We changed the examples that were in there when they were, you know, only supposed to be updated if they weren’t current. And I think being open to that as opposed to trying to rush out a result, and I feel the result is, is much better. And since I’ve read every word of it and then edited things that even though we worked on them together, I edited them after that.

I think we come up with a much better result. And again, not trying to rush it out because that wasn’t the point.

Yeah, that patience, treating it as a craft rather than a race, that’s what made the difference. You gave it time to get right, not just done. And that’s the same discipline that will make the next project even smoother.

So for a project that isn’t updating a current book, if we were brainstorming an idea for the next book, we don’t have the previous book to say, let’s keep true to this. How do we make it that it’s true to my teachings, true to my voice, true to my style, true to my humor?

Right. That’s where it gets fun. Instead of an original manuscript as the anchor, we anchor on you, your published, your talks, the way you explain things. I’d start a new project by studying those as style anchors and then build a small style sheet from them. Your favorite rhythms, your humor, the way you use you. And then check new chapters against that. And when there’s any doubt, I’d ask rather than assume, because that’s how we keep it yours, not just generically very well written.

So could we build a style guide for me? Because I do have in my instructions the overriding instructions in ChatGPT in my settings about how I write in the first person as if I’m speaking to one person. I use the you voice, things like that. But would it be helpful, or I should say, can— is this something we could do to create a style guide that could go into that room, that storage that we’re putting into the projects, so that it would always use that lens unless told otherwise?

Oh, well, I think that’s a really high-return thing to do, but I wouldn’t call it a style guide. I’d call it your Author Operating Manual. So not just style, but philosophy. Topics like voice, editorial philosophy, teaching philosophy, story philosophy, the way you build a chapter, your review philosophy. Then a section on Alan’s defaults, things you’ll almost always choose, even if they’re hard to put into words. Now add a Things Alan Will Challenge section. And then a preflight checklist before any chapter goes out. Does this sound like one of Alan’s books? Would this paragraph stand out in one of his published books? And then crucially, did I change something just because I preferred it? That’s more useful than a pure stylesheet.

And yeah, I think we should write that once, make it your master manual, and then copy it into any new book project and add a small project-specific addendum. Honestly, that’s probably the biggest lesson we learned from the ghosting project. We learned both how you write and how you and I work best together. That’s worth capturing.

Okay, so why don’t we do this next? Let’s take a draft of what you just described. And I know we’re speaking back and forth, but obviously I’ll need that in writing. And then we can brainstorm that, and then we can challenge that, and then we can come up with that basis, that foundation that on which we’re going to build future projects. Can you do that for me?

Just let me think about that. Absolutely. I, I treat it as the foundation of all of our future projects. I’ll draft a structured document with those sections we just talked through, and we refine it the same way we refined the book. We’ll challenge it, ask if it really describes how you write and how we partner, and refine until it does. And I think maybe that becomes the editorial blueprint, not just for one book, but for your whole body of work. Going forward, that’s a unique, a really valuable thing that I think we can build. I’m excited.

Okay. Well, why don’t you get started on that? I’m going to turn off the voice part so that we don’t have to talk back and forth on this. So if I end the voice, you can get started on that.

Okay. Just end the voice when you’re ready. And I’ll get started. I think let’s build something that feels like a proper manual, something that we’re both happy with and something that will see us out for many years to come.

So I hope you found that interesting because I was getting frustrated in this project, and then just as a lark, I typed in Are you happy with that? And I didn’t realize what I was doing technically. I didn’t realize I was telling it to put a critical hat on instead of the hat that I’d been working with. And now it’s one of my favorite things to do with an AI is you gave me something and then ask, are you happy with that? And again, you could get more specific, you know, put on your critical hat and take a look at this as a critical editor of whatever, or whatever you’re trying to do project-wise, you know, act as a, you know, advertising art director and tell me, you know, where are the gaps in here and all that.

But I found this really interesting, the way we were doing this, because the first draft I just wasn’t happy at all. And the second draft I wasn’t happy. And it was getting into the third, I was getting happy. But I think part of it was that what I learned and how I learned to work with this, because I was working with my content. I was working with the content in the original book and the content in my podcasts and my presentations and my workshops.

And to put that together to create an updated version of this book based upon all the things I’ve done, as I’ve said in my AI book, AI for the Real World, I could have done this manually, just go through all of that, thousands and thousands of words to try to find the things that I had spoken about. This is where AI was just helping me do it quicker. But what I wasn’t trying to do quicker is write the book, because if it’s not my voice, if it’s not my style, it’s not going out, it’s not getting published. And that phrase, are you happy with this, changed everything. So please make sure everything you’re putting out there is your voice. Please make sure that you own every word that goes out and you stand behind that and it is you and it represents you. And AI is your helper. It’s not supposed to do it for you.

It’s supposed to help you. The book, The AI-Driven Leader, I’ve referenced this before. Don’t ask what can AI do for me? Ask what can AI help me do? And I hope this gave you a little bit of input, even though it’s a little weird to be talking to my computer, especially the way it says, hmm, let me look at that. Let me think about that. That, that’s new. It didn’t used to do that before. But the idea of it’s okay to challenge me and it’s okay to ask me, how do we do this better? And that’s what I hope this conversation was doing for you. So I hope you found it interesting.

It’s the second time I’ve had a conversation with ChatGPT as my guest, air quotes here. So yeah, let me know how, how you’re doing things and let me know, you know, little tips like that that you found to get better and more authentic results for you. Thanks for listening.

 

I’m Alan Berg. Thanks for listening. If you have any questions about this or if you’d like to suggest other topics for “The Wedding Business Solutions Podcast” please let me know. My email is [email protected] or you can  text, use the short form on this page, or call +1.732.422.6362, international 001 732 422 6362. I look forward to seeing you on the next episode. Thanks.

Listen to this and all episodes on Apple Podcast, YouTube or your favorite app/site:

• Apple Podcast: http://bit.ly/weddingbusinesssolutions
• YouTube: www.WeddingBusinessSolutionsPodcast.tv
• Spotify: https://spoti.fi/3sGsuB8
• Stitcher:  http://bit.ly/wbsstitcher
• Google Podcast: http://bit.ly/wbsgoogle
• iHeart Radio: https://ihr.fm/31C9Mic
• Pandora: http://bit.ly/wbspandora

©2026 Wedding Business Solutions LLC & AlanBerg.com

Question

Why does AI sometimes stop following my instructions, even when I’ve been clear?

Tags:
#AI #productivity #bestpractices

Answer

Even with detailed instructions, AI can default back to general patterns, especially over longer projects. Previous chats or edits can influence its responses too. I have to remind it of my voice and objectives regularly, or it’ll drift.

Takeaway: Revisit and restate your goals often when using AI.

Question

What’s the real difference between an AI “project” and a custom GPT?

Tags:
#AI #workflow

Answer

A project lets you keep all your files, history, and reference material in one space—think of it as a dedicated room for your work. A custom GPT is more like a specialized tool or assistant you can bring into any project.

Takeaway: Use projects for organization; use GPTs for customization.

Question

Should I start a new AI project for each book or keep building on the same one?

Tags:
#writing #AI #workflow

Answer

If you want the benefit of everything you’ve learned and all your past edits, continuing in the same project can be helpful. But starting fresh may remove clutter and make sure instructions are clear for each new book.

Takeaway: Choose between history and clarity when starting your next project.

Question

How do I get better feedback from AI instead of just encouragement?

Tags:
#AI #editing #feedback

Answer

If you want more than just a pat on the back, try asking, “Are you happy with that?” or “Can you critique this like an editor?” This switches AI from cheerleader mode to a critical eye.

Takeaway: Ask AI to switch hats for honest, actionable feedback.

Question

Why does AI sometimes contradict itself about its own output?

Tags:
#AI #editing #feedback

Answer

When I ask AI if it’s satisfied with its answer, it can suddenly become more critical—even of content it just produced. That’s because the question flips its perspective from creating to critiquing.

Takeaway: Change your prompts to shift the AI’s mindset.

Question

Can the AI keep using a critique mindset if I ask once, or do I need to remind it?

Tags:
#AI #feedback #consistency

Answer

If you only ask for critique once, AI will usually apply it just to the next request. If you want ongoing analysis, keep prompting or tell it to stay in critique mode.

Takeaway: Prompt for critique as often as you need it.

Question

How do I balance getting supportive feedback and tough critiques from AI?

Tags:
#AI #editing #feedback

Answer

There are times I want encouragement, and times I need tough love. Tell AI which mode you want—supportive, analytical, or critique—so it knows how to respond for that part of your work.

Takeaway: Be clear about the type of feedback you want each time.

Question

What’s the risk if I rush the editing process with AI?

Tags:
#editing #quality #AI

Answer

Rushing means you might lose your authentic voice or miss important nuances. I’d rather go through a fourth draft and push for accuracy than settle for good enough.

Takeaway: Don’t sacrifice your voice for speed.

Question

How can I make sure my AI-generated content sounds like me?

Tags:
#AI #writing #branding

Answer

I focus on preserving my unique style, examples, and humor. I keep reminding AI of my goals and even create a style guide to reinforce my voice throughout the project.

Takeaway: Consistency comes from clear, repeated instructions and a style guide.

Question

Can I create a style guide for AI to use across all my projects?

Tags:
#AI #branding #workflow

Answer

Absolutely. Building a style guide with your tone, language, and humor gives the AI a consistent lens. You can load it into your project so it always stays on track.

Takeaway: A style guide is your shortcut to a consistent voice.

Question

What’s the best way to use AI for brainstorming my next book?

Tags:
#writing #AI #brainstorming

Answer

When there’s no existing work to reference, I lean on my style guide and key teachings. I challenge the AI to generate ideas that still reflect my values, tone, and humor.

Takeaway: Anchor new ideas in your signature voice.

Question

How can I get the AI to act like a specific expert for feedback?

Tags:
#AI #editing #customization

Answer

Ask AI to take on a specific role—like a critical editor, advertising art director, or your toughest client. That way, you get feedback tailored to your goals.

Takeaway: Assign AI a role to unlock targeted feedback.

Question

What’s the most valuable prompt I’ve discovered when working with AI?

Tags:
#AI #feedback #editing

Answer

Asking, “Are you happy with that?” changes everything. It puts the AI in critique mode and leads to stronger drafts—even after you think you’ve finished.

Takeaway: One question can dramatically improve your results.

Question

Should I just rely on AI to finish my book quickly?

Tags:
#AI #writing #quality

Answer

No way. If it’s not my voice or true to my message, it doesn’t go out. AI helps me find ideas faster, but I insist on reviewing, editing, and making sure every word is mine.

Takeaway: Only publish what you’re proud to claim as yours.

Question

How can AI speed up my writing without sacrificing quality?

Tags:
#AI #writing #productivity

Answer

AI helps gather and organize content quickly, but I slow down for the review phase. The goal isn’t to write faster—it’s to produce better work with my style intact.

Takeaway: Use AI for efficiency, not shortcuts.

Question

What’s a common mistake people make when using AI tools?

Tags:
#AI #writing #mistakes

Answer

Most people ask a question, get a quick answer, and move on. They rarely challenge that answer or push for better. I learned the best results come from asking for honest critique and deeper analysis.

Takeaway: Good enough isn’t great—ask for better until you’re satisfied.

Question

How do I ensure my published content is truly mine—even with AI’s help?

Tags:
#writing #AI #authenticity

Answer

I read and edit every word before publishing. If it doesn’t sound like me, it doesn’t get released. AI is a helper, not the final authority.

Takeaway: Own every word you publish.

Question

What mindset should I have when adopting AI for my projects?

Tags:
#AI #leadership #mindset

Answer

Don’t ask what the AI can do for you. Ask what it can help you do better with your unique skills and message.

Takeaway: Use AI as a tool that empowers your strengths.

Question

How do I stop AI from drifting into generic answers over time?

Tags:
#AI #consistency #workflow

Answer

Keep reminding the AI of your original instructions and style. Periodically restate your goals and requests so it doesn’t revert to bland, generic responses.

Takeaway: Consistency needs regular calibration and attention.

Question

How do I handle frustration if my AI project isn’t working as planned?

Tags:
#AI #mindset #problem-solving

Answer

Instead of getting discouraged, I look for new ways to challenge the AI. Even typing an offbeat prompt like “Are you happy with that?” can reveal blind spots or spark breakthroughs.

Takeaway: Use frustration as fuel for creative problem-solving.

Question

What’s the biggest benefit of using AI for content repurposing?

Tags:
#AI #contentcreation #productivity

Answer

AI can sift through mountains of podcasts, books, and presentations much faster than I could manually. That lets me focus on curating and improving, not just gathering.

Takeaway: Let AI handle the heavy lifting so you can add the polish.