The advent of ChatGPT has turbo-charged the debate around AI in copywriting, with many predicting it will rapidly lead to the demise of human copywriters. Is that true?
At first glance, ChatGPT looks like a huge opportunity. Words appear on the screen as if by magic. The content it produces sounds impressive, and when you run it through a plagiarism checker, you’ll find it’s unique.
If you can get a machine to generate copy in moments for free (or almost free), why would you pay a freelancer £387 a day or a full-time copywriter £53,123 a year to do the same thing and take a lot longer to do it?
I think there are several reasons why. Some of those reasons are temporary, and as AI develops, they’ll be overcome. Others are more permanent.
The currency and accuracy of the information generated by AI tools
The first limitation is that ChatGPT’s responses come from a model dating from early 2022. This is obviously problematic for any fast-moving field. Sooner rather than later, though, machine learning developments mean AI will be able to work using real-time information, so this will no longer be an issue.
A second limitation is that the accuracy of responses might not necessarily be there.
Stack Overflow, the Q&A site used by coders and programmers, has, for now, banned answers using ChatGPT. It says (the italics are theirs):
The primary problem is that while the answers which ChatGPT produces have a high rate of being incorrect, they typically look like they might be good …
There’s therefore a very real risk of misinformation being spread, especially when combined with the lack of real-time information.
A case in point: while researching this article I wanted to check how much ChatGPT knew about recent events, so I asked it about Nadhim Zahawi, who was sacked as Chairman of the UK’s Conservative Party on 29 January 2023 because he breached the ministerial code.
First, I asked ChatGPT to summarise why Nadhim Zahawi resigned. It told me:
Nadhim Zahawi resigned as the UK’s Vaccines Minister on January 31, 2023. The reason for his resignation is not publicly disclosed. However, it is not uncommon for political figures to resign for personal reasons or due to differences with the government’s policies. If you need more information, I suggest checking reliable news sources for any official statements or further details.
It sounded so confident I double-checked in case I’d missed a further development on the story. I hadn’t – Nadhim Zahawi held the post of Vaccines Minister in 2020 and 2021. He left the role when he became Education Secretary.
To give the tool a chance, I then asked it a more accurate question: can you summarise why Nadhim Zahawi was sacked? The second answer was equally confident – and equally wrong:
I’m not aware of Nadhim Zahawi being sacked as the UK’s Vaccines Minister. According to public information, he resigned from his position on January 31, 2023. The reason for his resignation has not been disclosed. If you have information to the contrary, I suggest checking reliable news sources for further details.
This is a fairly innocuous example of misinformation, but it’s very easy to see how dangerous it could be in the wrong hands.
The creativity question
If you want to coin a phrase, or express something beautifully, amusingly or eye-catchingly, a copywriter can give you what you need. You can also see the value of their linguistic creativity and dexterity when you consider that for more and more brands, their tone of voice is just as important as the way they look.
For many people, it’s this creativity that will mean copywriters will always beat AI. I certainly think that’s true at the moment. ChatGPT’s first answers do tend to have a similar, slightly verbose, tone.
But my sense is that AI-generated copy will rapidly advance in this area. While researching this article, I asked ChatGPT to write a blog post for me. It wrote in its usual style. I then asked it to rewrite it in a more informal tone. Then in a humorous tone. Then as a limerick. Then as a haiku. Every time, it managed it, in seconds.
The truth is, tone of voice is easily copied – for good reason. You don’t want to find yourself in a situation where only one person in your company can write in your tone of voice. Tone of voice guidelines show humans how to replicate a tone of voice. There’s no reason why they shouldn’t help AI replicate it too.
In a recent newsletter, Ann Handley gave instructions on how to come up with memorable copy that trumps AI. Here’s one example she gave:
ADD A SECOND SENSE. Show how something feels, tastes, smells, sounds… in addition to how it looks or what it is. Tap at least two senses.
I talk about this a lot. We don’t use it enough.
Example: Morning Brew taps this technique in an ad—an ad!—for Mugsy jeans:
“If [thighs] could talk, they’d be telling you how Mugsy jeans are the softest, comfiest jeans a set of thighs has ever called home.
“And if they could shout, these thighs would be screamin’ about the incredible stretch Mugsy jeans deliver, keeping the fit juuust right—never too baggy and never too tight.”
Talking thighs. Shouting thighs! That word just with three Us in the middle.You can feel AND hear AND see those comfy jeans giving your thighs plenty of space.
Much as it pains me to disagree with Ann Handley, the trouble is, if you can give instructions to a human, you can also give instructions to a machine.
You could argue that machines only copy, so they don’t have the power of originality. You can say the same of human copy too, though.
The trend to sound like Innocent Drinks was huge for a while, for example. Or, think about book titles. As the Financial Times points out:
[Julian Barnes’ novel] Flaubert’s Parrot launched a fashion for apostrophised titles – Foucault’s Pendulum, Lempriere’s Dictionary and dozens more (a trend lampooned in Bridget Jones’s Diary when she attends the launch of “Kafka’s Motorbike”).
In other words, I think we’re closer to AI being able to be linguistically creative than we might think.
But I also think there are different types of creativity. Linguistic creativity is one. Intellectual creativity is another. And it’s here where I think human copywriters will always be able to beat AI copy.
AI-generated content rehashes existing information
At the most basic level, AI-generated content is generic not specific (or, worse, is weirdly specific to the point of plagiarism being a real concern).
This is a problem if you want your copy to celebrate your company and show why prospective customers should choose it.
I’d draw an analogy with encounters I had when I worked as a web copywriter in a web design company in the mid-2000s. Clients occasionally asked me why I needed to interview them before I wrote the content for their website.
My answer was simple.
It would be very easy to write a website about a widget company. But I needed to know about the Acme Widget Company – what it stood for and why it was different to other widget companies. I could then write a website that clearly showed visitors why they should choose the Acme Widget Company rather than any of the other widget companies out there.
At the moment, web copy that’s generated by AI will be about a widget company not the Acme Widget Company. It won’t do the job it needs to do – giving your business a unique identity that helps it stand out in a competitive market.
But there’s an even bigger problem with this generic information.
Google has made it clear it will penalise content that’s AI-generated. Tools already exist to spot AI-generated content – including one from OpenAI, the same company that produced ChatGPT, to help teachers spot AI-generated homework.
Copywriters have always said that cheap copywriting is a false economy. AI-generated copy is also likely to be a false economy.
Where have all the ideas gone?
Because it rehashes existing information, there’s no new or unique thinking in AI-generated copy.
If we take the use of AI writing to its logical extreme, where everyone used it all the time, innovation and ideas would stop because we’d simply be restating things that had already been said.
For this reason, if you’re setting out your stall as an industry leader or an original thinker, copy and content that simply says things that are already out there, just in a different way, doesn’t cut it.
A lot of the work I do as a copywriter involves interviewing subject matter experts and customers to get fresh insights and unique takes on a topic. It also involves trawling review sites, forums and all the other places where my target audience is hanging out.
I take the information I gain from these processes and synthesise it into a fresh take, a fresh angle and create a compelling piece of copy that improves a company’s brand image and bottom line.
The actual writing of the copy, whether it’s 400 words or 4,000 words, is the last stage in the process and, arguably, the least arduous or complex. Companies don’t hire copywriters for that stage of the process (even though they might think they do). They hire them for the creativity and experience it takes to take a wealth of unique inputs and deliver an output that meets their precise objectives.
At the moment, AI can’t take a jumble of inputs that may or may not be important or relevant, pull out the salient points, and create an entirely fresh, new, coherent take on a topic.
What happens in the future if AI can work out what’s important and what isn’t?
But what happens if, in the future, that changes?
What if you could say to a machine: “Take these three interview recordings and these four forums. Work out the key points, common themes and interesting viewpoints. Align them with our product proposition. Then give me a landing page following the PAS framework that uses all this information.” What is the role of the copywriter then?
Well, the skills of a copywriter are still needed.
You still need someone to develop the questions that will tease out the information you need from the interviews. You still need someone to conduct the interviews – and ask follow-up questions when an answer opens up an interesting new theme.
You still need to research the places where your target audience hangs out.
You still need someone with a knowledge of psychology to see whether the AI output hits the right points in the right way – and to make (or ask the tool to make) the necessary amendments.
You still need a human on hand to check whether the tone of voice was right at a macro level (i.e. what’s the tone of voice that will resonate with our values and our customers’ values?) and at the micro level (i.e. does this piece of copy sound like us?).
As now, the value of a copywriter isn’t solely (or even mainly) in putting words down onto a page. It’s in all the effort that’s needed to distil all the inputs into that end result.
There’s a classic tale of an engineer called out to repair a machine that was critical to the running of a business. They walked up to the machine, took one look, then replaced a single washer. When the invoice arrived, it was for £400.50. When the finance manager queried why something that only took ten seconds to do cost so much, the engineer said simply: “It was 50p for the washer. It was £400 for all the years of experience that meant I knew exactly what the problem was and how to fix it so your business was back up and running in minutes.”
The value of copywriting has never been in the process of putting words on a page. In the wake of ChatGPT, it’s never been more important to remember that.
