Copywriter, content marketer and ghostwriter
Copywriter, content marketer and ghostwriter
Many copywriters out there are leery of AI. It’s coming for their jobs, it’s making us all stupider, it’s killing creativity. Or something.
I’m honestly not that fussed. Done right, AI helps copywriters – especially B2B marketing copywriters – get content out the door faster. We aren’t writing Shakespeare here, we’re writing stuff designed to get people to buy our stuff. I use AI in my work, and nobody has ever complained.
But it’s not a case of feeding the brief into the computer and sitting back. At least, not if you want to produce something that anyone wants to read. Let’s take a look at where AI can help you write great copy – and areas where it’s a massive chocolate teapot.
If you need someone to spitball an idea with, but everyone is busy working (or you work alone in your garden office like me), then AI can make a great sounding board. It mimics the back and forth between people well enough that the conversation feels real, and you can even speak directly to if if that works better for you. Because of this, it’s a powerful tool for developing an outline for a piece of content. Just be wary of its tendency not to challenge you on anything (see below).
You can also tell your AI the argument you would like to make, and ask it to find sources that support your position. You do need to check the data yourself, to make sure it hasn’t hallucinated anything (though that is getting less of an issue these days), but this is a quick and relatively easy way to ensure your content is well-researched.
This might sound strange, but putting the words down on the page isn’t the most exciting part of copywriting. All you’re really doing when you write a piece of content is taking the ideas you’ve already had and putting them in a format someone else can digest. If you have that idea properly fleshed out and articulated in your mind, AI can be brilliant at turning it into the finished piece of content.
The caveat here is that the outline needs to be strong. I recently used AI to write a piece for a client that was, I confess, pretty rubbish. I’d been in too much of a hurry to check that the oultine was solid – it wasn’t, and so the resulting piece was similarly bad. We got it fixed, of course, but it was a great reminder that bad inputs lead to bad outputs when it comes to AI.
And just in case anyone has started frothing at the mouth about creativity – relax. I recently read a fantastic thought piece by science writer Bill Hinchen positing that humans are no better at creativity than AI – or at the very least, we don’t tackle the challenge of being creative any differently. It’s well worth a read, even if it does fill you with existential dread and cause you to need a hug from a loved one.
People often point out, correctly, that LLMs like ChatGPT are, at their heart, fancy probability machines. They just choose the most likely words based on your request, based on all the previous works they’ve been trained on. That really isn’t any different to how a writer works. We are all influenced by what we’ve read; everything is, at its heart, derivative of everything that came before it.
Instead of worrying about whether the process is ‘creative’ enough, go and look at whether the final output does what you want it to do. I happened to hear a blues/soul reworking of Eminem’s “Without Me” on Spotify, which I then discovered was created by AI and was subsequently removed from Spotify. I haven’t hunted it out again because it’s obviously in contravention of copyright laws, but I would be lying if I said I didn’t still hum it. It slapped, as the kids say. The process behind it didn’t detract from its power as a piece of art, in my view. The same can be true of your content.
The only other thing to bear in mind is that length matters. Since LLMs are probability machines working word-by-word, the more words you ask them to string together, the greater the chance is that they will choose the wrong words. I find that blogs of up to 1,500 words or case studies of 1,000 words are fine. I tried to write a 5,000 word report using ChatGPT once, and it was a shit-show.
While AI is great for bouncing ideas off, you do need an idea to start with. AI isn’t (yet) that great at coming up with concepts you haven’t considered, or coming up with things that are fully baked.
If you’ve just asked ChatGPT or Claude to come up with some blog ideas for you, don’t expect genius, is what I’m saying. Maybe you don’t need genius, and you think that’s fine (looking at you, SEO marketers) but for those jobs where you want an idea that makes people go “woah”, you’ll still need a human in the loop to make sure the idea is impactful, relevant, and interesting.
I mentioned before that some LLMs – ChatGPT in particular – don’t challenge your ideas. They are unlikely to argue with you, will accept what you say as truth, and are generally more interested in agreeing with you than working with you. Notebook LLM is the only notable exception to this that I know of, thanks to my wife who is a power user of AI for teacher training. This means you need to tread carefully when using it to critique something.
You can ameliorate the effect by thinking carefully about your prompts. For instance, instead of asking it “do you agree that….” you could ask it “can you find any problems with [insert idea here]?” But the surest way to get a proper critique of your work is to give it to another human being, or let it rest for a while and then come back to it with your own brain.
The same goes for the outputs you get from your AI. Review its work like you would review that of an intern. Find the good in it, but don’t be blind to the limitations. It’s already saving you time by doing the writing – don’t skip the review phase.
Speaking of the review phase, in my experience AI is simply dreadful when it comes to fixing its own mistakes.
Sometimes it tries to preserve too hard to preserve the words it has written, and so doesn’t make the changes I ask for. Sometimes it makes the change, but in so doing screws something else up. Sometimes it claims it’s made the change when it hasn’t, and if you press it on the issue it has a sulk and crashes.
The only way around this is to be the editor yourself. Which, of course, relies on you being able to fix the issues you can spot – in other words, you need to be at least a bit good at writing.
This obviously makes using AI a bit more work than you might want, but there is good news: you can teach AI so that it maks fewer mistakes in future. Once you’ve edited your piece to your satisfaction, re-upload it to your AI and ask it to compare your version with its own. Once it’s highlighted the differences you can tell it to remember them and, next time you do a project like this, remember this feedback. Over time, you should get copy that’s nearer and nearer the mark.
It’s easy to think of writing as “putting words on a page”. But that’s like saying that cooking is “putting food on a plate”, or that marketing is “whipping up a brochure”. There is a lot of thinking, planning and preparation that goes on before the words make it onto the page, and that is all part of the skill of being a writer. AI can help with that, but it can’t do that for you. It can’t turn anyone into a great copywriter.
That’s why I’m not that fussed about using it, or that worried that it’s going to steal my job. It’s another tool I can use, and – like the keyboard, or Microsoft Word – it’s one that writers can use more effectively than non-writers (seriously I will challenge you to a Word-off).