I fed 700 articles into ChatGPT to find my "voice"
It's quite helpful
So recently, I created a custom ChatGPT filled with my old Medium articles (more than 700 of them)1.
Most of these are about creativity, writing, productivity, and more recently, note-taking.
Yes, I have too much time on my hands (hmmm…) or I just like to write (yes), but I wasn’t necessarily uploading them to have easy access to my archives or to find information.
I use it to analyze my tone of voice and style.
Those articles (you can check them here) are more social media-y than the ones here.
They aren’t quite clickbait, but they are more tip-driven than reflection-driven, unlike these.
Using ChatGPT for LinkedIn…
I’ve also been using ChatGPT to help me with LinkedIn comments.
I’ve posted very regularly on LinkedIn for the past four years and have cracked the code here and there, but I’m no guru.
I have about 9,500 followers, and to be honest, I thought I’d have more by now considering how much time I’ve put into it. (I feel that way about everywhere I write, though, which probably means I just need to keep going…)
But consistency is only one part of it. There’s also the part about expectations of a certain platform and which “templated” posts do better than others.
By templated, I mean popular formats.
Every medium has a format
LinkedIn has a format, just like Substack has a format, just like IG has a format, just like CNN has a format, just like an academic paper has a format, just like a novel has a format.
I’m usually loath to think too much about this (good writing is just good writing!), but we know things don’t work that way.
People come to content with certain expectations, and to “break” that takes a lot of work.
Usually, the best way is to play the game and then throw in a few “outlier” posts as an experiment.
Think of it like a Venn diagram—you definitely want parts of your writing to fit into the mold if you’re hoping to gain some sort of audience.
There are a million caveats to this, but honestly, this is not why I came here today.
I came here to talk about AI and using AI to write because I find it very useful.
But a quick story…
When I taught writing to community college students, we always had to have “the talk.”
“The talk” wasn’t that talk, but it was about sources and plagiarism.
Specifically, around Wikipedia and using it for research.
I would say that Wikipedia is a place to start, not the place to finish their research.
Wikipedia gives you a good background, a good summary, and sometimes a good list of sources to review and look into.
It’s a place to start, not a place to finish.
It’s the same with AI content. It’s the place to start, not the place to finish.
Give AI samples of your writing, do it
So with that in mind, generative AI gets a lot better once you give it samples of your writing.
It can then “think” in that voice, tone, and style, and also identify the topics that you regularly write about.
It can then generate a LinkedIn post, a blog article, or potentially a novel (haven’t tried this, but I should upload my old novels) in a certain style.
This sounds a bit disturbing at first, especially because your mind jumps to the wildest scenarios.
Like me uploading all of Abraham Lincoln’s speeches into one custom GPT, then asking it questions, and claiming Lincoln said it.
That’s a lark, and it quickly becomes old (unless you’re really into Lincoln).
It’s great at helping me identify what to write about next.
It’s another form of brainstorming. I can return to all of my past writings and then ask it for ideas or things to expand on.
Any good writer would review their old work for ideas anyway; this just speeds up the process.
It’s in “my voice,” and I have to establish that at first.
BUT IT’S NOT, you say. (I can hear you saying it). IT’S AI.
Well, what is “voice” or “style”? We all have to find it
And so we start emitting words, and then we write down those words, and that’s shaped by what our caregivers say (or don’t say) to us, how our teachers read and talk with us, and how they “edit” our work.
And then as writers, we read certain books and watch certain shows and movies, and then we gravitate towards the things we like and make decisions about what we don’t.
We are inhabiting our influences.
The most ambitious (or most sincere) of us go on to study under certain master teachers—those well-published and well-credentialed—who can then “critique” our voice and our styles, offering us tips about what they would do or how they would approach certain stories, all the while encouraging us to find “our voice.”
We’re already an amalgamation of our inputs.
Our voice is false; it’s not truly our own, and we even change it over time.
AI generative text, with its instant recall and critiques, can help us find our voices faster, rather than erasing them.
Try it and let me know how it goes.
Keep going-
Josh Spilker
I will do these Substack posts at some point. They seem very different and I often write in different styles for different audiences and didn’t want to mix it up