(Unapologetically, yes I used the dreaded em-dash above. And that is the right way to use it).

Last month I was looking to hire a copywriter. (Still am, btw). I did a "tagline test:” nothing fancy, just brief two copywriters so I get two directions, pick the winner, move on with my life like a normal person with a to-do list and a Slack inbox that won't quit.

So I ended up talking to the top chosen two. They have different backgrounds, exciting! The kind of setup where you'd expect, at minimum, two different thoughts. Maybe one goes conceptual while the other goes punchy. Maybe one overthinks it and the other nails it on instinct. Either way, you get range - and that's the whole point.

They both came back with the same line. WTH? Not similar, not "oh interesting, they landed in the same territory . . . “ in the words of Fred and George, "We're identical!" Same structure and rhythm, with that suspiciously polished energy that reads like it was workshopped within an inch of its life. Ninety percent word-for-word, like they'd copied off each other's papers in an exam room on opposite sides of the planet.

For a hot second I wondered if they'd found each other in some secret copywriter Discord and decided to gaslight me for sport. Then it clicked. I'd paid two strangers to think for me and they'd both outsourced the job to the same AI, delivering identical beige output with a strategy deck attached.

And once I saw it, I couldn't unsee it.

"It's not you, it's AI" ahemm, I meant ---> "It's not X, it's Y."

You know this pattern. You've definitely seen it 97,877 times this week alone, probably before your second coffee even kicked in. Once you clock it, it's like finding out about the Wilhelm scream . . . suddenly it's everywhere and you start wondering how you ever missed it.

Every ad, every LinkedIn post, every newsletter from billion-dollar household names down to your neighbor's drop shipping side hustle, I can keep going . . . All of them using the same rhetorical sleight of hand, hitting the same beats with the same confident energy, like a cover band that technically nails the notes but somehow misses the song.

Like they take the same masterclass, anddddd now AI is doing the homework for them.

Trapped in the Penrose stairs, I was doing my usual scroll (you know the one, where you tell yourself it's "research" while your calendar reminder gets snoozed for the third time) and I saw it. Big agency, the kind with offices in three countries and a client roster you'd recognize. They'd published an article about what kind of ads actually work right now.

We wrote that article last year. Same topic and angle, with a headline so close to ours I had a minor panic (wait, did we publish that?) and went digging through our drafts folder. We killed our version because it didn't pass our standards . . . it felt generic and surface-level, like everything else we were seeing at the time. So we spiked it and moved on with our lives. And then it showed up in my feed anyway, like hearing a melody you once hummed in the shower playing over the credits of someone else's film . . . that déjà vu hum you can't shake.

Pretty sure, the existence of our draft was never in their peripheral view. What happened is weirder than plagiarism. We've trained AI on a decade of marketing content, and now marketing content sounds like AI trained on itself. It's content ouroboros (the snake eating its own tail, except the snake is also writing your Q2 campaign and calling it "thought leadership").

Or as my friend Rimma Boshernitsan put it (she helps CEOs and CSOs shape the zeitgeist through DIALOGUE), with the kind of clarity I wish I'd had before writing this whole section:

Can we pleaseeeee. . . let’s minimize them. Or at least, if you use rhetorical flourishes and em-dashes, use them with intention.

Because I mean obviously we all use AI at this point. We're all on the same Penrose stairs just claiming we each took a different entrance, hitting the same vending machine at every landing that somehow only ever has one flavor. AI makes a great sparring partner, but most people are treating it like a ghostwriter. Because they want efficiency, and yet, efficiency is the death of anything worth remembering.

Here’s how our team does it: every piece of writing goes through at least three checks before it sees daylight. (and when I’m in my perfectionist mode, this draft can stay in draft mode for a while, it drives my PM crazy. . . but that’s a different challenge).

1️⃣ Teach AI To Write in Your Actual Voice.

This takes 10 minutes and you only have to do it once. Follow these steps:

Step 1: Record yourself talking about your topic for 5 minutes. Use voice memos on your phone. Use Wispr Flow. Don't script it, just talk like you're explaining it to a friend.

Step 2: Transcribe it. Use Otter, Rev, or whatever transcription tool you prefer. Keep the filler words, the tangents, all of it.

Step 3: Copy this prompt and paste it into Claude (or ChatGPT, or whatever you're using):

Step 4: Save the output as a custom instruction or skill in your AI tool so you don't have to do this every time.

Step 5: Still read everything out loud before you send it. AI will try to smooth out your rough edges. Put them back in.

He shared a list of phrases that mark you as someone who let the robots do your homework. The em-dashes, the "it's not X it's Y" reversals, the three-word dramatic beats, all of it. We added his whole list to our Claude skills. You should too.

p.s. Adding my most recent “ick” discoveries to this list: “quiet” (and “loud”) as in:

There, I did my inception. You’re going to start seeing “quiet” everywhere. You’re welcome.