Let me tell you about the moment I almost became part of the problem.
Last quarter, I was reviewing a batch of content from our team, and I noticed something unsettling: three different articles, written by three different people, all had the same cadence. The same throat-clearing openers. The same "Here's the thing" transitions. The same fake-profound endings about "the future being already here." It read like a choir singing in perfect, soulless unison.
We'd become an AI slop factory without realizing it.
Here's what nobody in the marketing leadership suite wants to admit: we're all using AI now. Every single one of us. The question isn't whether you're using it (you are, or your team is, or your agency is). The question is whether you're using it in a way that makes your content better, or whether you're just generating expensive noise that sounds like everyone else's expensive noise.
The 70% Problem
I've started calling it the 70% problem. AI can get you 70% of the way to a decent piece of content in about 10% of the time. That's genuinely remarkable. But that last 30%? That's where your brand lives. That's where your perspective lives. That's where the reason anyone should read your content instead of the 47 other articles on the same topic lives.
Most teams are shipping the 70% and calling it done. And readers can tell. They might not be able to articulate why something feels off, but they scroll past it. They don't share it. They certainly don't remember it.
Peter Yang's recent analysis identified over 20 distinct patterns that scream "AI wrote this": binary contrasts ("It's not X. It's Y."), colon reveals ("The best part: it learns."), dramatic fragments, importance puffery. Once you see these patterns, you can't unsee them. They're everywhere. They're probably in content your team published last week.
Our Actual Process (Warts and All)
So here's how we actually use AI for content at our shop, and I'm going to be honest about where it works and where it absolutely doesn't.
Research and synthesis: AI shines here. When I need to understand a new martech category or pull together perspectives on a trend, AI is genuinely useful. It can synthesize multiple sources, identify patterns, and give me a starting point that would have taken hours to assemble manually. This is the "boring automation parts" that Burk describes in his recent piece on avoiding slop.
First drafts: Proceed with extreme caution. We do use AI for first drafts sometimes, but never without what I call a "context dump." The AI needs to know our brand voice, our audience, our specific angle, and most importantly, what we're NOT saying. As Matt Giaro points out, AI hallucinates because it lacks context, not because you're bad at prompting. Every new chat is a blank slate that knows nothing about your positioning, your competitors, or your actual expertise.
The ideas themselves: This is where humans must stay in the driver's seat. Ask AI for "10 blog post ideas about B2B marketing" and you'll get the same 10 ideas everyone else gets. The generic is built into the system. Your ideas need to come from your actual work, your customer conversations, your data, your failures. Then you bring that idea to the AI, not the other way around.
The Editing Layer Nobody Wants to Do
Here's the uncomfortable truth: using AI well requires more editing, not less.
Ben Popper at Writer calls this the difference between treating AI as a "draft accelerator" versus an "infallible content machine." The teams that produce good AI-assisted content are the ones that treat the AI output as raw material, not finished product.
Our editing process now includes what I call a "slop check." We actively hunt for:
Weasel attribution. "Experts agree" and "studies show" without actual sources. If we can't link to the expert or the study, we cut it.
Synonym cycling. AI loves to call the same thing by three different names in three consecutive sentences to seem varied. It's not varied. It's confusing.
Fake profundity. Any sentence that sounds like it belongs on a motivational poster gets deleted. "The future isn't coming. It's already here." Cut. "In a world where change is the only constant." Cut. "This marks a pivotal moment." Cut.
The em-dash epidemic. AI scatters em-dashes everywhere like confetti. We've made it a rule: one per article, maximum, and only if truly necessary.
Building Your Own Slop Detector
The most valuable thing we've done is create what amounts to a house style guide specifically for AI editing. It lives in a shared doc that everyone on the content team can access and update. Every time someone catches a new AI tell, it goes in the doc.
Some of our rules:
- Never open with "In today's rapidly evolving landscape." (If I see this phrase one more time, I'm going to start a support group.)
- No "delve," "tapestry," "myriad," "plethora," or "multifaceted." These words have been ruined. They're AI tells now.
- Every claim needs a source we can actually link to. If AI made it up, we either find a real source or cut the claim.
- Read the final draft out loud. If it sounds like a Wikipedia article, it needs another pass.
The Real Competitive Advantage
Here's what I keep telling my team: in a world where everyone has access to the same AI tools, the differentiator isn't the AI. It's what you do with it. It's your actual expertise, your actual perspective, your actual voice.
AI is like having a very fast, very eager intern who has read everything on the internet but has never actually worked in your industry. That intern can be incredibly useful for certain tasks. But you wouldn't let that intern write your keynote speech or your most important customer communication without heavy supervision.
The companies that will win the content game aren't the ones producing the most AI-generated content. They're the ones using AI to amplify genuinely human insight, then doing the hard work of editing, refining, and injecting the perspective that no algorithm can replicate.
Marketing is still a team sport. AI just joined the roster. But it's not the captain, and it's definitely not the coach.
