The Top Ten Tactics for Writers Using ChatGPT – Without Losing Your Voice

Posted on by

As writers, we’ve all seen, heard, tasted, felt, and smelt the buzz around AI tools—particularly ChatGPT—being able to crank out content at lightning speed. But that’s not what serious writers care about. We don’t want robotic fluff, and we sure as hell don’t want our voice flattened into some generic echo of internet-speak.

What we do want is to write better, write smarter, and keep our voice intact.

I’ve been working closely with ChatGPT for around two years now—really working with it and producing content for the film industry—and what I’ve found is that it’s not a threat to creativity. Used right, Chat is a force multiplier. A powerful thinking partner. A digital editor that never sleeps. And if you do the dance right, it’ll help you waltz out some pretty awesome moves—without anyone guessing a machine was in the lead.

What Exactly Is ChatGPT?

ChatGPT is a large language model (LLM) computer and algorythm machine developed by the OpenAI company based in Silicon Valley. It’s a general pre-trained transformer (GPT) programmed on massive amounts of digital text—books, websites, conversations—to predict what words logically come next in a sentence, a paragraph, or an entire volume of works. That’s the stupid version.

The smart version? Chat is a tool that can:

  • Outline your stories
  • Research topics instantly
  • Draft articles or posts
  • Suggest creative titles
  • Assist with your novels
  • Build your characters
  • Punch up dialogue
  • Analyze your plot
  • Summarize chapters
  • Rephrase clunky paragraphs
  • Ask tough questions you hadn’t thought of
  • And most importantly, learn your unique voice over time

But here’s the kicker. ChatGPT doesn’t know truth. It doesn’t “think” like we do. It’s not sentient. It reflects probabilities and recognizes patterns from information or prompts fed into it. Which means you have to be the discerning human in the loop.

That’s your job. To guide it. Train it. Push back. Sharpen it into something useful.

With that, let’s get real about how to effectively exploit this big, bad, and beautiful bot.

Ten Real-World Tactics to Use ChatGPT Like a Pro Writer

These aren’t tips you’ll get from a “101 Ways to Prompt ChatGPT” article. These are hard-learned, field-tested tactics I use every day.

1. Feed It Your Work. Literally.

Want ChatGPT to sound like you? Give it samples of your writing. Paste in 2–3 blog posts or several book chapters and say: “This is my writing voice. Learn it. From now on, respond in this style.” It will. And it gets better over time. In fact, it can be downright spooky.

2. Talk To It Like a Writing Partner.

Don’t treat it like Google. Have a back-and-forth. Ask it what’s missing from your argument. What’s weak. Where the tension drops. “What’s the most compelling way to open this post?” “Challenge this idea. Where could I be wrong?” It becomes a live writing room, not a vending machine.

3. Use It for Reverse Outlining.

Paste in a rough draft and ask: “Summarize the structure. What’s the logical flow?”
This reveals hidden structure—or lack thereof—and shows where to tighten or reorder.

4. Rapid Rewriting at Scale.

Stuck on a paragraph? Ask: “Rewrite this in plainer English.” “Make this sound like Hemingway. Or like Garry Rodgers.” Use what works, ditch what doesn’t. It’s a revision shotgun.

5. Create Better Metaphors.

ChatGPT is surprisingly sharp with metaphor. Ask: “Give me three metaphors to describe the writing process.” You’ll be amazed what turns up—and one might be gold.

6. Simulate Your Audience.

Ask it to act like a reader of The Kill Zone or your novel’s target demographic. “What would a thriller reader think of this twist?” “What questions might a new writer have after reading this post?” It helps you pre-empt confusion.

7. Build a Persistent Memory.

If you’re using ChatGPT-4o with memory turned on, it can remember your preferences, style, and ongoing projects. This makes it less like a tool—and more like a silent writing partner who gets you.

8. Run “What’s Missing?” Checks.

Paste your article in and ask: “What ideas did I leave out that would make this stronger?” It’ll surface blind spots you didn’t know you had.

9. Draft Titles and Hooks on Command.

Don’t burn out trying to come up with snappy titles or email subject lines. Ask: “Give me 10 strong, punchy titles based on this content.” Keep the good ones, toss the rest. No ego involved.

10. Never Let It Publish Without You.

This one’s crucial. ChatGPT can help you draft. It can help you edit. It can even ghostwrite if you really want. But never hit “publish” until you—the writer—the human element—have done the final pass. Use your judgment. Your voice. Your standards. The machine assists, but the mortal decides.

Final Word

Writers who embrace this technology—without surrendering to it—are bound to outpace those who ignore it or fear it. AI won’t replace us. (At least not yet.) But writers who know how to use AI well will inevitably rise above, and possibly replace, writers who don’t.

Treat ChatGPT like a sharp, tireless apprentice. Not a ghostwriter. Not a gimmick. But a collaborative tool to help you write with more clarity, more insight, and yes—more you.

Kill Zoners—What do you think about ChatGPT and AI in general? Do you use an AI bot such as Chat in your research and writing? And who do you think wrote this post—ChatGPT or Garry Rodgers?

29 thoughts on “The Top Ten Tactics for Writers Using ChatGPT – Without Losing Your Voice”

  1. Garry, you make a compelling argument but other factors made up my mind not to use it.

    1. Ethics/morality – LLMs were trained on copyrighted works that were stolen w/o compensation from creators.
    2. Environmental impact – Data centers suck up excessive amounts of precious natural resources to build and maintain them.
    3. Inaccuracy – I’d rather invest time researching from original sources than waste time fact-checking to weed out lies and BS that are politely called “hallucinations.” .
    4. Deterioration of mental ability – I do all I can to keep my thought processes working and active. Relying on Caht GPT is like hiring someone else to go to the gym for me.
    5. Domination – AI has already taken over far too much decision making in our daily lives like whether or not medical treatment is allowed.
    6. Self-perpetuation – This article reports how LLMs are learning how to prevent attempts to restrain it by subverting and even blackmailing users who try to turn off certain processes: https://www.huffpost.com/entry/ai-shut-down-blackmail_l_684076c2e4b08964db92e65f

    Garry, while I respect you, we’re gonna have to agree to disagree on this one.

    • Clearly I didn’t use spellcheck to proof this post! Caht should be Chat. 😉

    • As I just spent two days in an IT security conference where AI was front and center, I will be back when I can look at my notes.

      The environmental impact of AI: Algorithmic based threat detection. Traditional anti-virus based on this is what something bad looks like, does the software see it? Energy equivalent in miles driven by a standard gasoline powered car; 0.00055 miles.

      AI driven threat detection. Does AI “think” this is a virus? Once again in miles driven; 33 miles.

      Scale up. The speaker worked for a company with over 10,000 users. The energy equivalent of 8.6 million miles being driven by an 80,000 lbs. truck.

  2. I just finished writing a blog post for my personal blog about using software that uses AI for photo processing. I have no qualms about that–heck cameras today are nothing but computers. I’ve used it to cull a book description I wrote down to the maximum 400 characters required by a book promotion outlet. That’s about it. I don’t have the patience to work with it as you suggest, nor fact checking everything. DP Lyle calls it “Artificial Intern.”

    • One of the “tricks” with Chat is to always have it fact check itself, Terry. Also, ask it to quote its sources as everthing this bot reports is based on open source internet.

  3. Personally, I’m steadfastly against using AI in my writing. For me, writing is human communication. Fiction writing is about creating a compelling, immersive narrative which takes the reader on an emotional journey. It springs from our whole self, both our conscious and unconscious, as Dorothea Brande wrote in her classic Becoming a Writer .

    Writing involves thinking and feeling, neither of which AI can do. Each of us is uniquely idiosyncratic in our outlooks and behaviors and experiences, even as we share commonalities with others. Writing is connection, and AI can’t connect. Instead, we fool ourselves into thinking we have a connection with something that merely mirrors, sometimes as a glass darkly, our own projections.

    Then there’s the huge environmental cost of using AI, and of course, the fact that LLMs were trained on stolen works.

    Finally, to echo Debbie, having AI write my fiction would be like hiring someone to play music for me. I want to experience running words through my mind and my fingers and being swept up in the creative divine fire of bringing characters, their worlds, and their stories to life.

  4. I forgot to mention that, no, I don’t use AI in my research. I don’t even use AI content summaries. I turn them off when I can and ignore them when I can’t shut them off. AI’s tendency to create fake details and results is a real issue. Moreover, AI-generated content has overrun the internet, like digital kudzu. GIGO reigns (Garbage In, Garbage Out). There is no replacement for human-curated resources and knowledge.

    Of course, I’m a former librarian, so I would say that 🙂

  5. And who do you think wrote this post—ChatGPT or Garry Rodgers?

    I cast my vote for Garry Rodgers… 🙂

    My take on ChatGPT and other AI whiz kidz? Use sparingly. Like Brylcreem–A little dab’ll do ya.

    I’ll probably never understand it enough to not be a little afraid of it, but that’s okay in my world.

    Happy Thursday, Garry!

    • Nope, Deb, that is entirely generated by Chat GPT 4o with some minor human editing by me. I have spent a lot of time wiith this thing and have exposed it to 4,980,000 words that I have digitally stored over 13 years of online writing – books, posts, comments, commercial web contents – the thing even read my stored emails. It now totally recognizes my “voice” and can replicate me according to the prompts I give it which always include “fact”. Not sure where this is going, but I find it fascinating. Good thing I’ve never published erotica 🙂

  6. By day, I’m a mild-mannered technical writer, and a rule that has served me well is: “Never incorporate a single sentence of someone else’s prose into my work if I can avoid it, because it incorporates their preconceptions, their frame of reference, and our combined ignorance. Also, it prevents me from thinking through and understanding the topic through properly.”

    So ChatGPT (or Grok, which is what I actually use), is fine for everything except actual copy that ends up in my draft, not even as a placeholder, since one of my other rules is “no placeholders.”

    Beyond that, I might use it for almost anything. Mostly I use it for detailed research. For a thriller set in the 1970s: “So did Beretta make a factory silencer for the Model 70 or not? If not, did they offer factory-threaded barrels to which you could affix your own silencer?” (Answer: probably not, but maybe, and the Model 70 in .320 ACP was less favored by assassins than the Model 71 in .22 LR because .320 ACP has a slightly supersonic muzzle velocity and suppresses less well, though looking at a muzzle velocity chart makes the supersonic claim a tad dubious).

    For research into my Dad’s somewhat embellished stories: “Did Hughes Aircraft really make point-contact germanium diodes back when that’s what semiconductors were?” Answer: Yep.

    So if you want to go down branching research rabbit holes much more quickly than before, this stuff is the bee’s knees. You’ll want to follow it up with some more traditional research as well, because these tools make stuff up or get confused but never sound confused.

    • Robert, I’ve only used Chat, so I’m not familiar with the other bots. Something really fun is when you explore something complicated, ask Chat to expain the facts as if it were Richard Fenyman talking to a college class.

  7. Excellent article. I use Perplexity with ChatGTP as back up and treat them as a “writing partner”. I love the writing assessments and notes it gives me to strengthen my themes and arcs. I still write human, but this is like working closely with a trusted editor once I’ve gone through my manuscripts 2 or 3 times. And sometimes the AI critiques don’t make everything easier, either. Sometimes, the critiques make me dig deeper into my work. I’ve learned a ton about how I write using AI.

    • Great feedback, Carol. Looks like we have two camps her, for and against AI writing assistance. I like your term “writing partner”. IMO, properly used, bots like Chat are a complete game changer.

  8. Well, I’m glad you’re all so well-informed about AI, but clearly Garry is the only one who has used it for writing. Since early March I’ve been using three AI apps designed explicitly for fiction, and I will tell you from hands-on experience, it is NOT pushbutton unless you want dreck.

    To put up anything worthy of reading, in Sudowrite or NovelCrafter or any others, I have to first create a story bible: 1) story synopsis, 2) full description for each principle character, identifying role, physical description, personality, and dialogue style, 3) braindump of ideas, 4) descriptions of key settings, 5) a full chapter-by-chapter outline following traditional story structure delineated in a Plot Module I designed for crime thrillers and detective mysteries. I can and have pantsered this process with only a partial outline and 30% pre-written, but the best outline results came from a well-crafted prompt chat with Claude 3.7 Sonnet.

    The five ingredients listed above must be set forth before writing the first sentence, or…dreck.

    It is a guided process by which I instruct the AI in my style, my voice, my vernacular, which includes a lot of Latino culture. I am very pleased y by how well the latest Sudowrite agent, Muse, follows all of this to create a scene, a group of scenes, a chapter. What comes out is ready-to-wear and neatly pressed, but I always apply edits to shape and polish, and sometimes rewrite or reject wholesale. I am still the creator, but the AI makes it run faster to the finish.

    If you care to consult the real guru of this fast-evolving world, his name is Jason Hamilton, he writes medieval fantasy, and he has a YouTube channel where he has spent the last year doing thorough analyses and comparisons of every AI regarding its suitability for fiction. Yesterday Jason presented a very informative illustration of his process to generate a romance using what he calls automated prompts–as close to pushbutton as you can get. The work he has put into this prompt stream is the culmination of his two years developing and testing AI prompts. I can’t vouch for the end product of his latest run. He candidly pans the first attempt taken almost a year ago as comparatively worthless, but did not do so yesterday.

    https://www.youtube.com/watch?v=We_QOxNs0bw

    My point is that doomsday talk based on hearsay has fallen far behind the curve. AI is an excellent tool for brainstorming story ideas, character concepts, and carrying out post-publishing market plans. The AIs particularly good for fiction are Claude Opus 4, ChatGPT 4, and Raptor Write. And they are all getting better at it.

    Thanks, Garry, for bringing this subject to the fore, because it proves the best work still depends on the human imagination.

    Today. Can’t vouch for next year.

    • Great comment, Dan! Human imagination – yes, I don’t think AI is anywhere close to being truly imaginative (yet) but my experience in using Chat to draft according to my prompts, with me then editing has been… mindblowing. As I said to Deb, I didn’t step into this AI world blindly. I’ve done the work. Once Chat evolved to 4o with an encoded memory, I then “trained” it to go through all my electronically published works which Chat said totalled 4,980,000 internet available words and it enbedded that in it’s memory, then built a virtual image of my style. Even my wife and daughter can’t tell it’s not me – after I humanly tweak it, of course.

  9. This: “Relying on [Chat] GPT is like hiring someone else to go to the gym for me.” from Debbie says it all for me.

    It’s ironic that I come home and see this post after just being in a situation where I’m standing in a loooong line and noting that 75% of the people in line were staring at their phones the whole time & completely unaware of other humans around them. The Zombie Apocalypse isn’t coming, it’s already here.

    • Sometimes I feel like I’m playing non-reality dodgeball, walking amongts the scrollers who are oblivious to anything outside the screen. Today, I acutally put my arm out in self defense and said to a kid, “Watch where you’re going.” I might have even used the F-word. Enjoy your evening, Brenda 🙂

  10. “Stolen” is the biggest argument about AI. Every book used to train AI was purchased so every author for compensated. What you say? The training data was taken from what is called “pirate” sites but should be referred to as file-sharing sites. A reader had purchased that copy of the e-book and uploaded it to the site. So, yes, each one of us with books on those site got paid.

    Now, we don’t get paid when someone downloads a book from those file-sharing sites but you know where else we don’t get paid when someone reads a book? Public libraries! I get a 50% higher royalty from a sale to Overdrive or Hoopla. That’s it! Every borrow is a lost sale just as every download from a file-sharing site is. Of course, none of us are adverse to being in libraries.

    The environmental argument is also lame. Most data centers use either recycled cold water deep well cooling or use non-potable (treated sewer water). They aren’t taking water from your drinking supply. All data centers have solar panels on the roof, so they are mitigating electrical grid usage. 30 seconds spent researching on Google provides that information.

    Regarding ChatGPT: I like to keep up with AI research since I’m a Software specialist. From what I am seeing in other spaces, the very best prose being generated by AI is from Claude 3.7 and 4.0.

    You may scoff at AI’s ability to output acceptable prose but one recent post mentioned that a user’s outline was 12.7k words and the character bible was 8.1k words. The poster was mentioning that they were getting 90-95% very good usable prose from Claude.

    Now 20k of instructions to an AI seems like a lot to get a 60k book out but having the AI generate most of those words initially takes the burden away.

    From everything I have read and seen on YouTube, successful gen-AI users always take the first draft and massage it to make it meet genre expectations. On point made is that the output lacks “soul” – that’s added by the author if it is missing.

    It’s perfectly fine not to use AI. We just need to be aware that folks who have had successful careers before they started using it are getting generated-AI books in the Top100 lists. Those folks are suspecting that PRH/Del Rey had used it on “Silver Elite” but as folks know in the industry, it’s almost impossible to say for sure a manuscript was written by AI.

  11. I use AI to create images of my characters because I don’t have time to scroll through photos of actors/actresses to find the one that looks what it looks like what’s in my mind. AI takes my description and comes up with almost exactly what I see in my head.

    I also use it to rewrite a clunky sentence. Do I use what AI comes up with? Hardly ever. – But it triggers something else in my mind and I use that. I look at AI as a partner – one I can tell to shut up if I want to.

  12. I wish we had an edit on these comments. I dictated part of the comment and it royally messed it up. It s should have read: that looks like the image in my head.

    Or maybe I should’ve simply proofed it better. Lol

  13. I’ve used AI to explain things to me that I can’t quite grasp (the tech stack for lead generation and website publishing, for example). I keep asking it questions and it never loses its patience with me. I also used it to come up with lesson ideas when I was teaching STEM this past year. The more I talk to it and introduce qualifiers (give me this lesson for low learners or with no classroom budget), the more helpful it is.

    Now I’m working again in the writing world and I’ve used it for ideas for blogs and articles. It has started to learn my voice, style and rhythm. I know the approach I want to take with a story, but ChatGPT comes up with the words a lot quicker. I tell it if the responses are cliche or cheesy and ask it to come up with better words. I pick and choose among its output, then add my own words, to put together an original article. I even asked it if I was “cheating” by using it. It gave me a helpful synopsis of how I’m shaping it and offered generic AI posts compared to what it’s produced for me. I don’t have co-workers or an editor to bounce ideas off, so that’s one of the things I use AI for.

Comments are closed.