Copywriting in an A.I. World

<< Back to the articles

Rain strategist and cofounder Noel Thevenet sat down with copywriter Bill Fahber to talk about how copywriting is being reshaped by the A.I revolution. Here are excerpts from their conversation.

Noel: So, is A.I. taking over copywriting?

Bill: In many ways, yes. And it should. Everyone should be using A.I., especially LLMs [Large Language Models]. But it has to be used wisely. A.I. can now do 80% of the work. But brands will live or die on how they master that final 20%. And that’s where most fall into traps. The tricky part is, this is all still new. Most clients won’t recognize the traps until they’ve already stumbled into them. And we’re seeing that happen a lot.

“A.I. can now do 80% of the work. But brands will live or die on how they master that final 20%.”

Bill Fahber – Head Copywriter, Rain

Noel: What kind of traps?

Bill: There are plenty. People joke about the overuse of em dashes, but that’s nothing compared to deeper structural flaws. A classic one is contrast framing – hyping up a benefit by dismissing a lesser alternative. For example: “It’s not just a vacation – it’s happiness on the beach.” Or “It’s more than a car. It’s a lifestyle on wheels.”

The technique itself isn’t bad. But an LLM can’t help overdoing it. It’ll spit out five of these over the span of a few hundred words. The result is heavy and repetitive copy. The same thing happens with choppy rhythms, repetitive patterns, and other strange obsessions. You can re-prompt all you want, but the model always slides back into these strange obsessions.

Noel: Studies show that consumers don’t like brands using A.I. How would they even know?

Bill: Any experienced writer can sniff out A.I. copy instantly. And I suspect frequent LLM users can too. You start to recognize its fingerprints. But the real issue isn’t using A.I. The real issue is depending on it entirely. A.I. is fast, cheap, and near-perfect with grammar. But it’s also bland. It can lack creativity and personality.

By design, it’s average. An LLM is trained on millions of data sets – what other people have already written. It doesn’t invent. It just recombines. So it can only ever be as good as what it has been fed.

Noel: Won’t the quality get better over time?

Bill: Maybe. But in the short term, it feels like the opposite is happening. LLMs are becoming inbred. With so much A.I.-generated content flooding the internet, the models are increasingly training on their own output. So, they’re literally inbreeding.

That means whatever they’re weak at, they’re only going to get weaker at. I imagine engineers will eventually find ways to refresh the training data, but for now the trend seems to be heading downhill.

Noel: You sound pessimistic for someone who champions A.I.

Bill: Oh no, I’m very optimistic. A.I. tools like LLMs are powerful. But they’re only as good as the people who use them. If you don’t know what quality, high-impact writing looks like, how can you tell when the machine is giving you junk? If you haven’t spent years A/B testing effective writing mechanics, how do you know what will perform best in direct-response copy? That’s why I always say, as a senior copywriter, A.I. is better in my hands than it is in yours.

Noel: So how has the role of senior copywriter changed?

Bill: The creative side hasn’t changed much. The clients who want real ideas still need real creatives. Even if we use A.I. to brainstorm, the copywriter has to know which sparks are worth fanning into fire.

What has really changed is the balance of the work. There’s less raw execution, more editorial strategy. Brands still need style guides and editorial frameworks. And if they want to make the most of LLMs to write in-house, they need senior-level specialists to show them how and help them build their tools.

Noel: You mean, custom GPTs?

Bill: Exactly! We can now create custom GPTs that follow brand guidelines, house style, and vocabulary. They can even act as watchdogs on CSR claims to help avoid greenwashing. That means you don’t always need a copywriter looking over your shoulder. Instead, you bring in a senior copywriter or editorial consultant to help you build your custom GPT: feed it, train it, test it, and show you how to use it.

And then, instead of distributing old-school editorial-guideline documents and hoping people follow them – which they rarely do – we can now work the guidelines into custom GPTs in-house. In theory, that means every piece of copy aligns automatically with a consistent brand voice.

Large enterprises are already developing their own A.I. and LLM platforms and working with writers to fine-tune them. But smaller companies and agencies that don’t mind using cloud-based systems can do the same with ChatGPT, Claude, or whichever LLM they’re comfortable with.

So much of this is still new, and we’re learning as we go. But these are exciting times!

Noel: No question about that. Thanks, Bill!