There’s something I should probably disclose about the articles on this site.
I use AI to write them.
Quite extensively, actually.
There. I’ve said it.
Run one through an AI detector and perhaps it will confidently announce that 90% of it was written by AI.
My response would probably be:
Get a life.
Because I think we’re starting to ask the wrong question about AI and authorship.
The interesting question isn’t:
Who typed the words?
It’s:
Where did the thought come from?
How these articles actually get written
Most of my articles start with an observation.
Something happens while I’m building software.
I have an annoying experience with an airport lounge app.
I notice something while using AI.
I remember something that happened 20 years ago.
Or I simply find myself thinking:
There’s something interesting here.
Quite often that observation first appears in a conversation with AI.
We talk about it.
If it feels like there’s an article hiding inside it, I don’t normally say:
Write me an article about this.
Instead, we have an interview.
One question at a time.
I’m asked what I mean.
I explain it.
Another question follows.
Sometimes I disagree with the premise of the question.
Sometimes I realise there’s another part of the argument I hadn’t considered.
Sometimes an example appears that changes the direction completely.
Some articles have ended up somewhere quite different from where I thought they were going when we started.
That’s why I like the interview process.
It isn’t simply gathering material for an article.
It’s part of the thinking.
Could I just ask AI to write the whole thing?
Of course.
I could type:
Write me a 2,000 word article arguing that software development in 2026 is about behaviours and outcomes.
A few seconds later I’d have an article.
It might even be good.
But I wouldn’t consider myself the author.
That’s why I don’t work that way.
If the AI invents the argument, chooses the examples, develops the position and writes the conclusion, then I haven’t authored anything.
I’ve commissioned some text.
There’s a difference.
Garbage in, garbage out
I’ve been writing about this recently in the context of software development, and apparently one of computing’s oldest rules follows me into writing too.
Garbage in, garbage out.
During these interviews, if the AI interprets something I’ve said incorrectly, I correct it.
If it starts developing an argument that isn’t what I believe, I stop it.
If it makes an assumption I don’t agree with, I say so.
That’s important.
Because AI is extraordinarily good at taking a bad premise and producing beautifully written prose around it.
Fluency doesn’t make the input correct.
My job throughout the interview is to make sure the observations, experiences, arguments and conclusions accurately represent what I think.
Only when we’ve done that do we write the article.
But I didn’t type this
No.
I didn’t.
And increasingly, I don’t think that matters.
I didn’t sit staring at a blank WordPress editor.
I didn’t spend an afternoon moving paragraphs around.
I didn’t agonise over whether one sentence should have seven words or nine.
AI did a substantial amount of that work.
But the article contains my observations.
My experiences.
My statements.
My opinions.
And quite often, my actual words.
The thing I didn’t do was physically type most of them into the finished document.
Does that mean I didn’t write it?
I don’t think it does.
What about my voice?
There’s another part to this process.
Before we started writing these articles, I provided a document describing how I write.
There are things I like.
Things I don’t like.
Certain rhythms.
Certain ways I structure an argument.
A bit of British humour.
And punctuation habits that AI seems otherwise determined to inflict on everybody.
If you’ve noticed a suspicious absence of em dashes around here, that’s not an accident.
The document doesn’t provide the ideas.
It provides constraints around how those ideas should be expressed.
So there are really two inputs.
The interviews establish what I think.
The style establishes how I communicate.
AI works inside both.
Sometimes AI helps me discover what I think
This is probably the bit that makes authorship more complicated.
I don’t always arrive with a completely formed argument.
Sometimes I just have an observation.
The interview develops it.
A question makes me think about something differently.
I answer.
The answer produces another question.
I disagree.
We go somewhere else.
Eventually an argument emerges.
So I can’t honestly claim that AI is merely a sophisticated typewriter.
It isn’t.
It’s participating in the thinking process.
But participation isn’t authorship.
The important thing, for me, is that I retain editorial authority throughout that process.
Nothing ends up here simply because an AI decided it sounded convincing.
I have to believe it.
I have to be prepared to defend it.
And if you disagree with it, you’re disagreeing with me.
Provenance of thought
Perhaps that’s the distinction I’ve been looking for.
We’ve become very interested in detecting the provenance of prose.
Was this sentence generated by a human?
Was this paragraph generated by an AI?
What percentage of this document looks machine written?
I’m much more interested in the provenance of thought.
Where did the observation originate?
Whose experience is being described?
Who developed the argument?
Who challenged it?
Who corrected it?
Who decided what stayed?
Who is willing to put their name underneath it?
For the articles on this site, I can trace that provenance.
Often all the way back to some ridiculous thing that happened to me that morning.
The photographs matter too
There’s another deliberate choice I’ve made with this site.
Wherever possible, I use my own photographs.
Sometimes I’ll take a photograph specifically for an article.
Sometimes I’ll remember something in my photo library that fits.
I have used AI generated imagery occasionally, usually when time made it convenient.
But increasingly I avoid it.
I learned that lesson while marketing my music.
After seeing enough perfectly polished AI generated imagery, I reached the point where I couldn’t take any more of it.
The AI slop brigade had won.
There’s something about using a photograph I actually took that matters to me.
It doesn’t need to be perfect.
It was there.
I was there.
It has provenance.
And perhaps that’s the same thing I’m looking for in the writing.
Not proof that AI never touched it.
Proof that there’s a human experience underneath it.
I already have a word for this
While writing this article, I realised something slightly embarrassing.
I’ve already spent an entire article describing exactly this process.
Except I was talking about software development.
I’ve stopped writing most of my own code.
Instead, I describe what I’m trying to achieve.
I provide the context.
I question the plan.
I watch what the AI does.
I challenge decisions I don’t agree with.
I test the result.
I correct the direction.
And when I’m satisfied with the outcome, I ship it.
I’ve started calling that orchestration.
And that’s exactly what I’m doing here.
The medium has changed.
The process hasn’t.
With software, I’m not outsourcing the engineering.
I’m orchestrating the implementation.
With writing, I’m not outsourcing the thinking.
I’m orchestrating its expression.
That might be the distinction I’ve been looking for all along.
Authorship as orchestration
For most of history, authorship and production have been tightly coupled.
If I wrote an article, I probably typed the sentences.
If I wrote software, I probably typed the code.
So we naturally associated authorship with production.
AI breaks that relationship.
I can now be responsible for something without personally producing every component of it.
My role moves upwards.
I provide the intent.
I provide the experience.
I establish the constraints.
I question the output.
I correct the direction.
I decide when it’s finished.
And ultimately, I take responsibility for the outcome.
That’s orchestration.
It doesn’t mean I can put my name on anything an AI generates.
Quite the opposite.
If I ask AI to invent an argument and then publish it because I happen to agree with what came back, I haven’t orchestrated much of anything.
The provenance matters.
The thinking matters.
The involvement matters.
And the responsibility matters.
So, did AI write this?
Yes.
And so did I.
AI interviewed me.
AI challenged some of my thinking.
AI helped develop the argument.
AI constructed much of the finished prose.
I reviewed it.
I corrected it.
The observations are mine.
The experiences are mine.
The opinions are mine.
The editorial decisions are mine.
And my name goes at the top because I’m prepared to stand behind what it says.
If I’d simply typed “write me an article about AI authorship” and published whatever came back, I wouldn’t feel comfortable doing that.
That’s the line for me.
I don’t think counting keystrokes is going to tell us very much about authorship anymore.
Ask where the thought came from.
Ask whether the person whose name is on it can explain it.
Challenge the argument.
Disagree with it.
That’s far more interesting than feeding it into an AI detector.
And if your only contribution is to tell me that a machine thinks a machine helped write it…
Well.
Get a life.