100% Written by Me. 100% Written With AI.
I use AI when I create content. By default, actually. And, increasingly, happily and with full conviction.
Even as I see that sentence materialize on my screen, I feel a faint urge to defend myself. Yes, I use AI, but the ideas are mine. Yes, it helps me write, but I rewrite. Yes, I use ChatGPT, but I could write before generative AI existed. Somewhere along the way I started to wonder why I was behaving as though I’d been caught.
Back at school, writing came easily. I generally earned top marks for creative writing; I wrote lyrics for our band and discovered the small thrill of other people responding to words I had written; and even when our language teacher terrified the class with modern poetry, I was happy to have a go. Perhaps the shameless confidence of adolescence simply let me go down the slippery slope of writing without ever touching the handbrake.
Then business writing came along, where words committed to paper could later be quoted back at you. As long as it had a clear job to do, I could still produce presentations, proposals and internal arguments without much hesitation. But ask me to write a public piece that simply said, “this is what I think and this is why,” and the machinery changed. Every sentence looked permanent, every claim wanted another qualification, and every paragraph seemed to apply for planning permission before I could build the next one.
I tried the recommended remedies. Ann Handley’s TUFD told me to get the rough thing down before making it good. John Steinbeck’s second rule urges us to write freely and rapidly, then repair later. Stephen King distinguishes between writing with the door closed and rewriting with the door open. All three, in different ways, reveal one of the keys to better writing: allowing the bad draft to exist before asking it to be good. And that was exactly my problem. Even the rough draft felt like an MVP, and in my head the V mattered far more than the M. It had to feel viable before I let it exist. Once my words started hitting Word, the handbrake came back on.
My first, almost successful, workaround came years before AI. I started dictating, because speaking had never suffered from the same problem. Give me a stage and an idea I understand, and language tends to follow the thought. Connections appear, arguments move around, and sometimes a phrase leaves my mouth in better shape than I knew it had been in my head.
Dictation got me moving, but it rarely got me all the way to Draft One. Somewhere mid-draft, the handbrake came back on: I started repairing sentences before I had finished drafting the thought.
Then AI entered, and it did something dictation alone couldn’t: it helped me keep moving long enough to reach Draft One. It did not remove the doubt; it moved it later in the process. The exchange was fast and fluid enough for me to stay with the thought instead of stopping to repair every sentence as I went. In effect, AI finally helped me do what Ann Handley, Steinbeck and King had all been pointing towards: let the imperfect draft exist before trying to make it good.
“It did not remove the doubt; it moved it later in the process.”
My process is not throwing rough ingredients into a machine and waiting for polished prose. I may start with a dictated argument, a few paragraphs I wrote, or an idea I can see clearly in its unfinished shape. I ask what argument AI hears, where the logic jumps, or what I am assuming without saying. Sometimes it finds the missing link. Sometimes it confidently finds the wrong one, which can be just as useful because explaining why forces me to become clearer.
From there we work back and forth. I add context, it proposes a structure, I reject it, it questions a distinction I think matters, I explain why, it offers alternatives, and I dislike most of them. By now, AI and I are almost like an old couple: we disagree wholeheartedly on a regular basis. Somewhere in that mildly argumentative exchange, the thought usually gets better.
Sometimes it hands me back my own thought with the hedging removed. Sometimes I reject what it gives me entirely. Either way, the conversation keeps me moving. After more experimentation than I care to count, I have landed on a conclusion that still feels odd: AI is the best writing partner I have ever had.
AI and I are almost like an old couple: we disagree wholeheartedly on a regular basis.
That does not make it the author. If I ask it to manufacture expertise I do not have, invent experiences I never had or formulate opinions I never bothered to form, something has gone badly wrong. The problem is not that AI touches the work. The problem would begin if I disappeared from it.
“The problem is not that AI touches the work. The problem would begin if I disappeared from it.”
AI slop is real. But let’s be honest: human slop existed long before ChatGPT. AI did not invent mediocre content, it just made mediocre content nearly frictionless to produce. So the opposite of AI slop is not necessarily human-only content. What interests me more is human-empowered AI. I bring experience, judgement, curiosity, taste, questions and stubbornness. The machine brings capabilities I lack. Sometimes the combination beats what either generic AI or my old solo process would have produced.
And if my name is on the result, I remain responsible for it. We spend a lot of energy policing who used AI, how much, and whether they should confess to it. But when the question is whether the resulting content deserves anyone’s attention, I think we may be asking the wrong question. We keep questioning the writer about the purity of the process, while the reader has a much simpler encounter: is this worth my attention?
Readers do not calculate the purity of your production process. They stop, scroll past, question, disagree, share, remember, or carry a thought somewhere unexpected. Sometimes the point I considered central barely registers, while a small observation halfway through starts the conversation. That is useful, not because I want to optimise every sentence for clicks, but because listening tells me what stayed unclear and what mattered to readers that I had underestimated.
“Readers do not calculate the purity of your production process.”
Which, I suppose, is my actual confession, if I still need one. I sat down thinking I had to admit that I use AI. Having written it out, I’m mostly left wondering why I thought it needed admitting at all. The partnership gets better as I get better at directing it, resisting it and listening to readers. Dictation taught me how to release the handbrake. AI helped me stay with the thought once I started moving.
And if you still want a percentage, this is the only one I can honestly give you:
100% written by me. 100% written with AI.
With my name attached, so you know exactly where to send your agreement, disagreement, questions or objections.