There’s a 45 per cent chance this op-ed was written by AI
PR Emma Divers was trained to write using the exact technique now being flagged by AI detectors. She says the AI crackdown risks becoming a witch hunt
According to one AI detector, there’s a 45 per cent chance a machine wrote this. The most generous assessment from another gave me 15 per cent – ‘AI polished’, apparently.
The problem is that I wrote every single word.
It’s a strange experience being told by a robot that your writing sounds like it was written by another robot. Especially when you’ve spent the last decade being trained to write in exactly the way AI detectors seem to hate.
I’m a PR. I write for a living.
From press releases to opinion pieces, my client gives me a story to tell and 500 words to do it. My job is to write something interesting enough to hold your attention, simple enough to understand and short enough that you’ll actually finish it. All while keeping in mind that the average reading age in the UK is somewhere between 9 and 11 years old.
All of that means writing in short sentences. Clear structure. Minimal jargon. The occasional one-line paragraph to drag the reader towards the point.
And yes, I use the rule of three. Beginning, middle, end. Blood, sweat and tears. You can pry the rule of three from my cold.dead.hands.
Unfortunately, those are all now things that will land you in AI jail.
My agency, like many others, uses AI detection tools to make sure we’re not contributing to the ever-growing mountain of AI slop. And that’s led me to the ridiculous situation of running my own work through AI detectors and then actively making it worse so it scores as more “human”.
Over the past few weeks I’ve found myself deliberately introducing awkward phrasing, janky grammar and stylistic quirks simply to convince a piece of software that I am not a robot.
Think about how absurd that is.
One colleague recently redrafted a perfectly good opinion piece four separate times just to satisfy an AI detector.
The original told a deeply personal story. It was thoughtful, well-structured and moving. The detector absolutely hated it. (Thoughtful, by the way, is a word flagged by AI detectors as a ‘significantly AI word’).
To score 100 per cent human, she had to break up the logical flow and remove the one-line paragraphs that landed the emotional punches.
I eventually made her put the changes back in. Then I spent the rest of the week worrying that a journalist would accuse her of not writing the piece herself and blacklist us from their inboxes.
At this point, AI detectors owe me a sick line.
Don’t get me wrong – I have huge sympathy for publications like Morning Wire putting their foot down on AI submissions. They’re right to care. They’re right to protect standards. They’re right to worry about an internet drowning in AI slop. And it’s only because of the mountain of badly generated writing that I’m in this position.
But the starting point can’t be an AI-powered AI detector. Especially when many of those same tools offer a handy “humanise” button designed to help people evade detection in the first place.
The answer needs to be the foundations of what makes any PR or journalist good at their job.
Does the writing tell a story? Does it add something new to the conversation? Does it reveal a perspective, experience or argument that couldn’t exist without a human behind it? Does it make the reader think something they hadn’t thought before? And at the end of all that – is it well written?
Because despite all the hype, we’re not yet living in a world where AI can consistently replicate genuine insight, personality or lived experience. People instinctively hate AI content. But we don’t hate it because of em-dashes and perfect grammar. We hate it because it boils down to lots of pretty words strung together in a way that says absolutely nothing of note.
And if we become so obsessed with spotting robots that we start penalising clear, effective writing, we’ll all end up where I find myself – being forced to write badly just to prove that I’m real.
And at that point, AI won’t have lowered our standards. We’ll have done that ourselves.
Emma Divers is a senior associate director at PR agency PLMR