The Em-Dash Defence League

I've been thinking about what we win and what we lose when we allow AI into our writing space.

It's a thorny topic. When I was doing my celebrant training, we may as well have been asked to sign an oath in blood: thou shalt not use AI to write ceremonies. The underlying message was clear enough: this work is meant to be authored, not assembled.

I took that steer seriously. It wasn't hard: I love writing, and I'm certainly not ready to hand over all the fun to an algorithm. And yet in spite of this, my mentor had been reviewing one of my ceremonies one day and mentioned to me that although it didn’t read like I'd used AI, her daughter had pointed out that my use of em-dashes was “a red flag.” It seems there are apparently now entire corners of the internet devoted to spotting the tell-tale signs. Certain phrases raise suspicion. Overly polished paragraph structures. An enthusiasm for bullet points and short sentences. And yes, one particularly damning giveaway: the em-dash.

Reader, I was affronted. Because I love an em-dash. I love dashes generally, but I particularly love the em-dash. I use them with wild abandon — sometimes responsibly, sometimes like a woman scattering birdseed in a public park. I even know the keyboard shortcut: Shift + Option + Dash (on a MacBook, obviously. I am not a complete Philistine).

For what it’s worth, the em-dash has been around for centuries. Writers like Emily Dickinson, Virginia Woolf, James Joyce and Mary Shelley used it long before anyone started scanning punctuation for signs of artificial intelligence. In fact, when I was having lunch with my firstborn last week (a talented writer herself and also no stranger to the em-dash), she told me that if you feed excerpts from Dracula into some AI detection tools, they'll tell you it's been something like 95% AI-generated simply because the writing is so structured and consistent. Which is both hilarious and horrible — rather like accusing Victorian literature of suspiciously good grammar.

AI detection tools don’t actually “detect AI.” They estimate how statistically predictable text is. That means they can mistake clarity, rhythm, or formal structure for something machine-like. They are notoriously unreliable, leading us to believe that good writing is suspicious simply because it is well written.

I think that's pretty tragic. As I mentioned earlier, I LOVE writing. It’s not just my bread and butter: it’s my life breath. I was writing long before ChatGPT appeared on the scene: I won my first writing competition aged nine. The prize was a tin of Cadbury’s chocolate goodies, and I was beside myself with happiness. Since then I have published novels, written speeches, ceremonies, articles, scripts, technical content, and deeply personal pieces for families navigating some of the most significant moments of their lives. My relationship with the em-dash is longstanding and entirely authentic. If AI has now adopted it as a stylistic quirk, I would like the record to show that I got there first.

And for the avoidance of doubt, every ceremony I write is completely unique and entirely my own work — albeit shaped, enriched, and made infinitely better by the stories, memories, and input shared by my couples and families. I don’t use templates. I don’t churn out generic scripts with the names swapped around. I write ceremonies the way I’ve always written everything: by listening carefully, noticing the important details, by immersing myself in every story and finding the thread that makes it unmistakably yours.

(Also, my ceremonies are exquisitely punctuated. If this matters to you, you are very much my kind of person.)

That said, I’m not anti-AI. I think it’s an extraordinary tool when used thoughtfully. It’s useful for sanity-checking a piece of writing. For spotting when you've used the same word thirty-seven times in three paragraphs. For helping untangle a sentence that has wandered off into the wilderness and can no longer remember where it started. Occasionally, I’ll ask it for a witty one-liner to introduce a blog post, although inevitably I'll go ahead and rewrite it anyway. Still, it will have sparked the idea. Credit where credit's due.

What I don’t think AI should do is replace the messy, deeply human business of storytelling.

The reason ceremonies matter — the reason stories matter — is because they come from lived experience. From noticing things. From loving people. From grief and absurdity and joy and years of paying attention. However good ChatGPT might be — however much it can search phenomenal amounts of information online and sound like your best friend when it's doling out advice — it doesn't have lived experience. Sure, it can help polish a sentence. But it can't sit beside a grieving family and notice the way they all laugh at exactly the same memory. It can't watch the way a couple will reach for one another's hand when they're telling you what they love best about each other. It can't know what it feels like to stand on the deck of a boat on Scotland’s west coast with your husband beside you, your dogs asleep at your feet, and the sunset turning the water molten gold.

And even though it uses them with gay abandon, it definitely can't appreciate the quiet satisfaction of deploying a perfectly judged em-dash. Of course it can't: it's a machine, remember.

For a while after I found out about this rather alarming interpretation of the em-dash, I stopped using it. Reader, I capitulated. I caved to the pressure of trying to be seen as unique, even though I know I am. And then I thought, hold on a sec: why should I compromise the quality of my written work over a misconception? I should be CHAMPIONING the em-dash, not bailing on it.

(Actually, I didn't really think 'hold on a sec', but manners prevent me from writing what I really thought.)

Tools are not the enemy of creativity. They’re just tools.

The magic still comes from the human being holding them.

Even if she does have what some may see as an unhealthy obsession with punctuation.