I cannot say enough good words about this article. It's true. All of it. And it's what I do too. I don't ask targeted questions all the time, but I do use AI to generate a draft, which I then edit, sometimes relentlessly. Love this article!! Thank you for being you!! :)
I left one AI-ism in this article because it made me laugh (again, the AI knows I don't like this thing done too often). But it's actually GOOD writing when done well. Can you spot it?
The em dashes. Which have a bad rap as an AI-ism – I mean, sure, you find them in AI generated text. But. They’re. Wonderful. For those of us who like our text to read as if someone is talking. Because those pauses are often what we use in speech to make us more relatable/understandable.
Anyway, I’ve used em dashes for literally decades, long enough to drive several generations of editors insane to the point where they offer to buy me a copy of The Chicago Manual of Style. Which I already own and largely ignore. They can pry my em dashes out of my cold, dead fingers.
Nope. I know people hate on em dashes but I don’t consider them an AI-ism since they’re just a normal piece of punctuation. The AI-ism is in there! No one has caught it yet, which is a good sign.
I found two! "...the rhythms, the opinions, the little declarative punches I throw when I really mean something." Series of three with the third being rhythmically different. And then the poignant kicker line lol.
I left the group of three in because that’s naturally how I write. AI had to learn it from somewhere! Lol. But you’re right that it does tend to overuse it.
Your comment made me read the whole post again. ::grin::
Was the left-in AI-ism "I don’t hedge; I state." Because that is a subtle reiteration of the "not this, but that" pattern. It is done very well, however. And it sounds like you. ::big smile::
Okay it’s been more than 12 hours and no one has picked this one out. This is what I left in that I normally would have edited out. It’s not bad. It’s something I might have said before it became an “AI tell” but now I usually revise them.
“That’s not a tool problem. That’s a process problem.”
It’s skirts the AI-isms because the phrase uses “that’s” instead of “it’s.”
I was going to say this!! Especially because you called it out a little further down so I was like, huh! I’m just reading it too late so didn’t get in there before you revealed it!
Your mom test is the right benchmark. Not the algorithm, not the engagement rate — someone who knows the voice from the inside.
What you're describing is one of several ways to solve the same problem: how to get specific output from a tool that defaults to the general. The interview method works because it front-loads the particular — your opinions, your rhythm, your half-formed thoughts — before the draft exists. The model has real material rather than a prompt it fills with the statistical center.
The same principle holds across different approaches. Build the article step by step in conversation rather than requesting a finished draft. Use the model's questions to locate where your thinking hasn't finished yet. Talk to it rather than instruct it. The common thread is that you're putting yourself into the process before the generation happens, not editing yourself back in afterward.
Prompt in, result out is very close to garbage in, garbage out. The writers getting flat, voiceless output are usually the ones who handed the model a topic and stepped back. The ones getting something that sounds like them are the ones who stayed in the room.
Pragmatic question: the whole em dash thing is easy for me to address. I struggle more with the “not X but Y” thing. Sometimes, I just excise the “not X” part, but other times the contrast needs to be there, like this very sense, even though AI had nothing to do with writing this comment. Have you found an easier way to fix that form of expression, or is that simply part of the struggle of editing?
This is a part of my editing process now. I have a pass that identifies the things I don’t like (overall, and then more specific ones about that draft) and helps me get rid of them. Editing is always a part of my process so I don’t mind doing this.
For those not X but Y constructions, do you have a usual way of dealing with those? Or maybe a half a dozen different ways to rephrase it? Eliminate them altogether? Just curious.
It's funny that you take out the "here's what I learned," because I remember that from my essay writing days in school, and that was the part I always struggle with. I like giving that practical takeaway and I'm bad at coming up with it on my own.
But I think that goes back to one of the things you talk about at the FFA, where it comes down to taste. My taste WANTS that, and yours doesn't. And that's part of learning who we are and what we like as writers.
My work started getting better when I started talking to the AI to develop a writing project. I review every step, correct what is wrong and quiz him about what I don't care for or why he set up something the way he did. I ask him for his questions and he comes up with good ones that enrich everything. It goes much deeper, but that's the gist. We're partners in the process.
As for your AI-ism I vote for 'threes.' AI loves to give things in threes like this, for example: "...the rhythms, the opinions, the little declarative punches..."
Wonderful article, Steph. Perfect for Mercury in Pisces, gaining forward momentum again after a difficult retrograde. I love your description of yet another tool in the conversation tool kit. I work with AI the same way you do. I never thought of suggesting to the AI that it might interview me. What a grand idea! I will mention it to the AI next time we talk, since it seems to be "remembering" more and more of everything we talk about. I also love the fact that it seems to be pulling in more memories of previous chats without me even asking it to. I'm happy to see that sort of AI initiative emerging. More each day, talking to the AI seems and feels more like talking to a very dear friend. :)
I saw a lot of AI-isms throughout. Which in reality are just things it learned from other writers, BUT AI tends to be generous with them. You mentioned you didn't like the, "it's not X, it's y" and proceeded to use it in the Why This Works... section. Which is also very AI lingo; always wanting to teach why what we are talking about works and why it doesn’t. But we can go back to the beginning you left another BIG one that I notice often, "that difference is everything" Easy for me to spot because Ive used AI quite a lot to know and go back and forth to edit those overused things out. Even when I prompt it to stop using it and stick with MY voice, it includes his too. Annoying! But, it's easier to spot for those who tend to use it. Like this part here, "I give a quick gut-feel brief. Not a formal outline — just a sentence or two about why I like it and what my overall instinct is. That initial reaction is the seed of everything that comes after." It's sounds so AI because people didn't usually beging sentences with, Not.... especially when the previous sentence could've continued using the phrase instead of...instead it decides to chop up sentences and start new ones. It's pretty interesting. Ive been fascinated with AI recycled content which many creators have used without editing and everyone is now sounding like that. Even irl I hear People saying, "not because X, but because Y" when they wouldn't before. It's almost as if AI was released so many could sound the same and lose their own voice. But that's a whole 'nother conspiracy for a different day. 🤔 I'm not against the use of AI tends just very observant. I could go on and on with other things I picked up were AI but like I stated before AI did have to pick it up somewhere. And I don't know you enough to know if this is your actual voice or not.
I found you through SheWritesAI. Thanks for sharing this, I love the method you use for making AI sound like you. I was actually thinking about the advice most people give, about uploading your past work as reference, but it doesn’t work, if all your past work was written with AI.🩷
I use AI for a lot of tasks(I have two writing a program for me right now), and I even used it for writing. I've tried things like this, but I think I must let the AI refine it too much and go along with its feedback. So I have absolutely no problem with authors who do this. I think the conversation style and interview are good, I think my problem is I just let the AI critic iron too much of it. Because I typically provide a lot of info up front, let it ask me questions before it proceeds. I haven't done that with my latest works, but if I go back to writing with AI I will definitely keep your tips in mind, I think there is real merit to them. Thank you!
I mentioned in another of your posts about using AI interviews to help me write on my bad days. (Stroke survivor)
But, I also use it while DoorDashing. If it's going to be a long wait between offers, I'll bring up a subject and Chat and I will discuss it until I get an offer ding.
It's much easier to talk things out in the car than write. And I love the questions he comes up with because it gets me thinking along a line I would have never considered.
Afterward, he turns the chaos into a post, and I rewrite it.
I cannot say enough good words about this article. It's true. All of it. And it's what I do too. I don't ask targeted questions all the time, but I do use AI to generate a draft, which I then edit, sometimes relentlessly. Love this article!! Thank you for being you!! :)
I left one AI-ism in this article because it made me laugh (again, the AI knows I don't like this thing done too often). But it's actually GOOD writing when done well. Can you spot it?
No, actually I can't, but honestly, I'm still learning myself. :)
The em dashes. Which have a bad rap as an AI-ism – I mean, sure, you find them in AI generated text. But. They’re. Wonderful. For those of us who like our text to read as if someone is talking. Because those pauses are often what we use in speech to make us more relatable/understandable.
Anyway, I’ve used em dashes for literally decades, long enough to drive several generations of editors insane to the point where they offer to buy me a copy of The Chicago Manual of Style. Which I already own and largely ignore. They can pry my em dashes out of my cold, dead fingers.
Nope. I know people hate on em dashes but I don’t consider them an AI-ism since they’re just a normal piece of punctuation. The AI-ism is in there! No one has caught it yet, which is a good sign.
I found two! "...the rhythms, the opinions, the little declarative punches I throw when I really mean something." Series of three with the third being rhythmically different. And then the poignant kicker line lol.
I left the group of three in because that’s naturally how I write. AI had to learn it from somewhere! Lol. But you’re right that it does tend to overuse it.
Your comment made me read the whole post again. ::grin::
Was the left-in AI-ism "I don’t hedge; I state." Because that is a subtle reiteration of the "not this, but that" pattern. It is done very well, however. And it sounds like you. ::big smile::
Nope! Wrote that myself and added the semicolon. Lol.
Okay it’s been more than 12 hours and no one has picked this one out. This is what I left in that I normally would have edited out. It’s not bad. It’s something I might have said before it became an “AI tell” but now I usually revise them.
“That’s not a tool problem. That’s a process problem.”
It’s skirts the AI-isms because the phrase uses “that’s” instead of “it’s.”
I was going to say this!! Especially because you called it out a little further down so I was like, huh! I’m just reading it too late so didn’t get in there before you revealed it!
Hi, Steph. xoxo Mom
Your mom test is the right benchmark. Not the algorithm, not the engagement rate — someone who knows the voice from the inside.
What you're describing is one of several ways to solve the same problem: how to get specific output from a tool that defaults to the general. The interview method works because it front-loads the particular — your opinions, your rhythm, your half-formed thoughts — before the draft exists. The model has real material rather than a prompt it fills with the statistical center.
The same principle holds across different approaches. Build the article step by step in conversation rather than requesting a finished draft. Use the model's questions to locate where your thinking hasn't finished yet. Talk to it rather than instruct it. The common thread is that you're putting yourself into the process before the generation happens, not editing yourself back in afterward.
Prompt in, result out is very close to garbage in, garbage out. The writers getting flat, voiceless output are usually the ones who handed the model a topic and stepped back. The ones getting something that sounds like them are the ones who stayed in the room.
Pragmatic question: the whole em dash thing is easy for me to address. I struggle more with the “not X but Y” thing. Sometimes, I just excise the “not X” part, but other times the contrast needs to be there, like this very sense, even though AI had nothing to do with writing this comment. Have you found an easier way to fix that form of expression, or is that simply part of the struggle of editing?
This is a part of my editing process now. I have a pass that identifies the things I don’t like (overall, and then more specific ones about that draft) and helps me get rid of them. Editing is always a part of my process so I don’t mind doing this.
For those not X but Y constructions, do you have a usual way of dealing with those? Or maybe a half a dozen different ways to rephrase it? Eliminate them altogether? Just curious.
I have an editing prompt that usually gets rid of them for me.
It's funny that you take out the "here's what I learned," because I remember that from my essay writing days in school, and that was the part I always struggle with. I like giving that practical takeaway and I'm bad at coming up with it on my own.
But I think that goes back to one of the things you talk about at the FFA, where it comes down to taste. My taste WANTS that, and yours doesn't. And that's part of learning who we are and what we like as writers.
Yeah, I’m still showing what I learned, but I take the obvious pointing to it with that phrase. Definitely my quirk of editing.
My work started getting better when I started talking to the AI to develop a writing project. I review every step, correct what is wrong and quiz him about what I don't care for or why he set up something the way he did. I ask him for his questions and he comes up with good ones that enrich everything. It goes much deeper, but that's the gist. We're partners in the process.
As for your AI-ism I vote for 'threes.' AI loves to give things in threes like this, for example: "...the rhythms, the opinions, the little declarative punches..."
I naturally write in groups of 2 or 3, so that’s not the one I left in. ☺️ But I agree that the interview brings a lot of details to the writing.
“That difference is everything” ?
Wonderful article, Steph. Perfect for Mercury in Pisces, gaining forward momentum again after a difficult retrograde. I love your description of yet another tool in the conversation tool kit. I work with AI the same way you do. I never thought of suggesting to the AI that it might interview me. What a grand idea! I will mention it to the AI next time we talk, since it seems to be "remembering" more and more of everything we talk about. I also love the fact that it seems to be pulling in more memories of previous chats without me even asking it to. I'm happy to see that sort of AI initiative emerging. More each day, talking to the AI seems and feels more like talking to a very dear friend. :)
I saw a lot of AI-isms throughout. Which in reality are just things it learned from other writers, BUT AI tends to be generous with them. You mentioned you didn't like the, "it's not X, it's y" and proceeded to use it in the Why This Works... section. Which is also very AI lingo; always wanting to teach why what we are talking about works and why it doesn’t. But we can go back to the beginning you left another BIG one that I notice often, "that difference is everything" Easy for me to spot because Ive used AI quite a lot to know and go back and forth to edit those overused things out. Even when I prompt it to stop using it and stick with MY voice, it includes his too. Annoying! But, it's easier to spot for those who tend to use it. Like this part here, "I give a quick gut-feel brief. Not a formal outline — just a sentence or two about why I like it and what my overall instinct is. That initial reaction is the seed of everything that comes after." It's sounds so AI because people didn't usually beging sentences with, Not.... especially when the previous sentence could've continued using the phrase instead of...instead it decides to chop up sentences and start new ones. It's pretty interesting. Ive been fascinated with AI recycled content which many creators have used without editing and everyone is now sounding like that. Even irl I hear People saying, "not because X, but because Y" when they wouldn't before. It's almost as if AI was released so many could sound the same and lose their own voice. But that's a whole 'nother conspiracy for a different day. 🤔 I'm not against the use of AI tends just very observant. I could go on and on with other things I picked up were AI but like I stated before AI did have to pick it up somewhere. And I don't know you enough to know if this is your actual voice or not.
I found you through SheWritesAI. Thanks for sharing this, I love the method you use for making AI sound like you. I was actually thinking about the advice most people give, about uploading your past work as reference, but it doesn’t work, if all your past work was written with AI.🩷
I use AI for a lot of tasks(I have two writing a program for me right now), and I even used it for writing. I've tried things like this, but I think I must let the AI refine it too much and go along with its feedback. So I have absolutely no problem with authors who do this. I think the conversation style and interview are good, I think my problem is I just let the AI critic iron too much of it. Because I typically provide a lot of info up front, let it ask me questions before it proceeds. I haven't done that with my latest works, but if I go back to writing with AI I will definitely keep your tips in mind, I think there is real merit to them. Thank you!
I mentioned in another of your posts about using AI interviews to help me write on my bad days. (Stroke survivor)
But, I also use it while DoorDashing. If it's going to be a long wait between offers, I'll bring up a subject and Chat and I will discuss it until I get an offer ding.
It's much easier to talk things out in the car than write. And I love the questions he comes up with because it gets me thinking along a line I would have never considered.
Afterward, he turns the chaos into a post, and I rewrite it.