I’ve started skipping over things when I read. Someone posts on LinkedIn, it’s perfectly readable, and somewhere around “I came expecting one thing and left with another” or “the thing that landed with me” or “I needed to sit with that for a while”, I find my eyes have already slid to the next item. It’s subconscious. It isn’t a decision. I don’t want to skip it but by the time I notice I’ve done it, I’ve moved on. And I’m not sure if I like myself doing that…
Why? Mostly, these are not lazy AI-slop posts. Sometimes, I know the person well, I know they’re not getting a machine to think for them. What they’ve probably done is exactly what I do: taken their own notes and thoughts and asked ChatGPT or Claude to help them draft something, or tidy up a rough version, tighten the structure, make it read more cleanly. There is nothing wrong with that.
And yet I find myself skipping over it anyway.
Why does skipping this AI assisted content matter?
We’re told AI democratises writing, and in one sense it clearly does. More people can get from a rough thought to a clear, publishable paragraph, especially if they’re time-poor, less confident, or writing in a second language. That’s an important leveller. But producing fluent-sounding writing, and anyone actually paying attention to it, are not the same thing.
When enough writing carries the same AI ‘texture’, readers start filtering on the texture itself:
- The tidy reversal (“*I came to give the advice. I left having taken it.*”).
- The symmetrical little list where the third item is the emotional turn (“*Cleaner code, faster shipping, and – ultimately – happier humans.*”).
- The reflective sign-off (“*Anyway. Just something I’ve been sitting with. Keen to hear how others see it.*”).
- Perhaps the biggest tell, the “it’s not X, it’s Y” false inversion (“*This isn’t a job. It’s a calling.” / “It’s not a setback, it’s a setup.*”).
- Not to forget the endless litany of leverages, sharpens, pivotals etc…
- And I am not going anywhere near the controversial pandora’s box that is the em–dash!
These have started to signal – consciously or otherwise – that we should filter out the content. This filter is crude, and very often wrong, but it happens. AI-cleaned prose is like the 2010s version of 90s clipart or Comic Sans font – competent, inoffensive, but becoming almost invisible.
The latest form of gatekeeping
Writing was always a barrier that acted as a gatekeeper to being taken seriously – better ‘standard’ English, a more educated and well-read style, more confidence and more practice writing for a professional audience – all of that already shaped whose ideas travelled. And rarely fairly – we have always penalised people with great ideas who don’t express them as well as the next person’s well-written mediocrity (yes academic journals I am looking at you!).
What’s happening now is that we have a different form of gate that rewards different skills. It’s not about how well you can write but how well you can instruct your favourite AI tool to disguise its built-in proclivities, so your writing doesn’t sound like it was written by AI.
What’s worse is that the markers aren’t even reliable. AI didn’t invent these phrases. It learned them from us, then repeated them so relentlessly that the original human versions now sound synthetic. A while ago I read a post by someone I know, genuinely good and genuinely interesting, and midway I read “I needed to sit with that for a minute.” I was about to share it and caught myself… Unfairly, as for all I know she wrote every word herself and has always talked exactly like that, and just happens to have a voice that overlaps with the flourishes that AI has made unacceptable.
And the truly daft part is how unfairly this sorts people. Take two writers. The first has done all the real thinking – their own ideas, proper notes, something genuinely worth your time – and simply got AI to turn it into clean prose. Eighty or ninety percent of what matters is theirs, but it reads machine-made, so we glide straight past it. The second lobbed a lazy one-line prompt into ChatGPT, thought about almost nothing, then either edited it well or had a style guide that sounds human – so ninety percent of it is really the machine, and we read every word. We’re judging them in precisely the wrong direction: the one who thought least gets heard, the one who thought most gets filtered out.
What am I doing in my own writing..?
My own response is, honestly, symptom management. I keep adding to a list of banned words, phrases and rhetorical moves I won’t use, not because they’re inherently bad or wrong but because they now signal the wrong thing.
But this now means others decide whether or not to take me seriously based on my knowledge of ‘AI-isms’, my skill at getting Claude to avoid them, and my ability to do a good final, edit – rather than anything about the merit of what I have to say!
That’s a new kind of advantage – not better ideas or deeper experience, but better “AI editorial hygiene”. I like to think I am pretty good at it, but I don’t really think it should have any impact on who reads what I write..!
So while I’m reasonably convinced by the “AI democratises writing” argument – I’m much less sure it democratises being taken seriously. I am making an effort not to skip over good content solely due to AI triggers but… I fear it may be a lost cause!
An ironic postscript I noticed while drafting this: A few times now, I’ve caught myself saying ‘banned AI-isms’ out loud – on a zoom call hearing myself say “let’s see where something lands” for example (an expression I’m pretty sure I never ever used pre-AI). I’ve almost certainly picked it up from Claude, which of course stole it from human writing in the first place. So the phrase goes from ordinary, to an AI-tell, to banned, to quietly reabsorbed by people who never used it before. At this point, does it become human – and allowed – again!? Are we on a never-ending merry-go-round of policing what we can and can’t say. Well I guess it too wouldn’t be the first time!
