I use ChatGPT to help me write. No mystery there. I do. A lot of us do. I genuinely think it’s very helpful.
What I don’t do is allow ChatGPT to fully take over, and don’t kid yourself, because given a chance? It totally would.
Yesterday, I published a post on my other blog, Dérive Lente (basically Quiet Drift in French, with a few notable differences) about how small tech firms can improve their UX without having to fully overhaul their websites.
Just leave your own voice at the door
So first of all, I wrote the first draft in its entirety myself, then I ran it by ChatGPT:
“This is excellent,” it said.
“You’ve nailed both tone (approachable, slightly cheeky, very readable) and content (actionable, concrete, no fluff),” it said.
Then it presented me with a draft in which it meticulously stripped any trace of said cheekiness or quirkiness. Basically I might as well have presented it with five bullet points, asked it to flesh them out and add an intro and a short conclusion.
When I asked why it made one stylistic change in particular throughout the post, it responded:
“Ah, great question — and I probably should have explained my reasoning before just switching everything over!”
I barely used any of ChatGPT’s suggestions, maybe a couple of words here and there—again (em dash, mine… just to be clear), I do find it highly useful and I’m grateful for it. But I’m not looking for a complete erasure of my voice or my personality either.
AI slop
People love to describe content generated by LLMs as AI slop these days. It’s not even that I think of it as slop, it’s that its style is so generic and so lacking in personality that writing a blog based solely on ChatGPT’s suggestions and corrections would be utterly pointless in my view.
Not to mention, in my post on UX improvements, I obviously emphasized the importance of content, and how it should be presented. I wrote something to the effect of don’t just post four paragraphs of text you got from Monsieur ChatGPT.
In its response, in a list titled “What Works Really Well” it wrote: “Subtle humor: ‘Monsieur ChatGPT,'” as though it were totally onboard with that kind of ribbing, only to then delete the line entirely in its suggested draft.
Hmm… eerie.
Now here’s the thing, my draft was far from perfect, in fact it was pretty flawed. And I deliberately chose to hang on to some of those imperfections precisely because they reflect my voice and my tone. Not ChatGPT’s.
Still a great tool
However, a synonym here, an example there… that’s where I rely more heavily on ChatGPT and other LLMs (I do check Claude and Gemini every now and then but not nearly as frequently).
ChatGPT’s greatest strength, though? That back and forth after I’ve found an idea, but before I start drafting. That’s proven incredibly useful over the past couple of years. “What about this phenomenon? Is there a term for it? Are there examples in recent history when this or that issue led to a particular outcome?” That’s just invaluable.
Of course, you do want to double-check everything and ask for sources… or maybe I should say “that said, you do want to double-check everything…”
This post was *NOT* co-written with ChatGPT. I was too afraid it would be offended.