AI Didn't Replace My Copywriters. Here's What Actually Happened
I spent the last year doing something that probably makes me sound insane: I ran 12 A/B tests pitting AI-generated ad copy against human copywriters across six different industries. Dental marketing, real estate, restaurants, podcasts, SaaS tools, sports teams. We tested it all.
The AI enthusiasts thought I'd prove that machines are taking over. Human creatives thought I'd show that AI is garbage.
Instead, I found something way more interesting: you're all wrong. And you're all right.
Here's What Actually Happened
We kept everything identical. Same images, same targeting, same budget. The only variable was who wrote the copy: GPT-4, a custom Meta Ads GPT, Meta's AI suggestions, or a human copywriter.
The results?
Humans won 9 out of 12 tests for cost per result and 5 out of 12 for click-through rates.
But here's the kicker. AI didn't suck. In some scenarios, it actually crushed it. The podcast campaign? AI got a 1.18% CTR while the human copy only hit 0.61%.
For a dental marketing agency, switching to human copy dropped their cost per lead from $7.09 to $4.50. That's a 36% reduction. Real money.
Why Humans Still Win More Often
Three things humans bring that AI can't quite replicate yet:
Emotional context. Restaurant and dental marketing needed that local, trust-building element. AI can copy emotion. Humans understand it.
Cultural nuance. High-stakes decisions like real estate investment need the human touch. People buy from people they trust.
Strategic storytelling. I used to front a rock band in Japan. I learned early that technical ability isn't enough. You need to connect. Same applies here.
Where AI Actually Shines
Let's be honest. AI isn't trash. It's just different.
AI generates 10 variations in 30 seconds. It produces content at scale without burning out your team. It handles the painful part of copywriting: the blank page, the shitty first draft, the 10 versions you need for testing.
AI isn't replacing copywriters. It's replacing the grunt work.
The Approach That Actually Works
Here's the method we use now:
AI generates 5-10 variations (takes 90 seconds)
Human picks the best 2-3
Human adds the emotional layer. The stuff that makes you stop scrolling
Test both when it matters
Learn from the data
Think about it like this: AI is your band's recording equipment. Essential. Makes you sound better. Speeds things up.
But it's not the performance. You're still the performer. The audience still wants to connect with you, not your gear.
What This Means for You
Content creators who refuse to use AI are like musicians who refused electric guitars. Creators who only use AI are like DJs who never learned music theory.
Neither extreme wins. The middle path does.
After analyzing thousands in ad spend across 12 campaigns, here's my confession: I was hoping for a clear winner. AI or human. Black or white.
Instead, the truth is way more interesting: we need both.
AI makes us faster. Humans make us better. The creators who figure out how to use both, who let AI handle the drafts and use their own brains for the connections, those are the ones who'll dominate.
Because at the end of the day, people still want to work with people.
They just want those people to be efficient as hell.
Guest Author: Jason Hunt is the co-founder of Merged Media and author of Drop The Mic Marketing. His full research on AI vs human copywriting was published in the Journal of Brand Strategy. Connect with him on LinkedIn or check out his favourite AI tools.