3 things I learned about ChatGPT
All I wanted to see was a few more ‘U’s and less ‘Zees’ but I guess ChatGPT ‘favors’ US spelling and I need to work on writing my prompts.
Is ChatGPT ‘the death’ of copywriting? Well, according to an answer from the model itself it’s “a tool that can assist humans in their work, but it’s not a substitute for human creativity and expertise”.
As a Senior Copywriter at Sticky, I thought I’d test the idea by sticking to the morbid theme and using an unusual prompt: ‘Write a short About Us page for a funeral director in a controversial tone of voice’. Here’s what I learned while tinkering with the brand christened Eternal Rest by ChatGPT:
1. It’s quite literal
I asked ChatGPT to write in a controversial tone of voice and I saw ‘controversial’ in the copy, I asked it to write something with a humorous tone of voice and saw the word ‘humorous’ in the copy. It’s hardly a revelation but ChatGPT loves the literal.
I wasn’t ready for how literal it could be when I asked for a humorous version of the About Us page using British English. I thought I’d see a slightly surreal twist to the text and a ‘u’ added to ‘humor’, instead, I was given a barrage of Blighty stereotypes.
Eternal Rest became Eternal Chuckles as part of its first rebrand and then mentions of Monty Python and Shakespeare suddenly appeared:
“Fancy a funeral theme that reflects your loved one’s passions and eccentricities? We’ve got you covered! From Monty Python-themed processions to a touch of Shakespearean drama, our creative team will tickle your funny bone while preserving the essence of your dear departed.”
ChatGPT
My favourite sentence was definitely: “Enjoy a cuppa with some scrumptious scones as we share tales and anecdotes that will have you laughing till your cheeks turn as red as a London bus.” All I wanted to see was a few more ‘U’s and less ‘Zees’ but I guess ChatGPT ‘favors’ US spelling and I need to work on writing my prompts.
When the New York Post put the tools’ stereotyping to the test, it labelled Brits as ‘uptight’ and ‘tea-loving’ and made sure to mention: “another stereotype about British people is that they have bad teeth”. Charming.
2. It will lie to you
When I asked ChatGPT to add some stats to the About Us page it came up with the bold: “Did you know that 87% of our clients felt a sense of comfort and healing through laughter during their loved ones’ ceremonies?”
Comic relief has its place in a funeral and is probably a useful grieving tool but the idea that Eternal Rest/Chuckles has invented its own clients and their satisfaction rates is slightly worrying.
OpenAI’s CEO Sam Altman has said that ChatGPT will “confidently state things as if they were facts that are entirely made up”. The brains behind the bot call it the ‘hallucinations problem’ and it could be more problematic than we think. “I’m particularly worried that these models could be used for large-scale disinformation,” he added.
In a 2023 study, 697 people read 220 tweets including accurate and false information written by humans and ChatGPT-3. Topics included climate change, vaccine safety, flat Earth theories and other controversial subjects. When the group guessed which tweets were true, Chat GPT proved to be slightly better at deceiving readers than humans.
There is some hope for us though – humans proved better at working out the accuracy of information than GPT-3. Take that, AI.
3. It’s not a search engine
‘What would a good funeral director do?’, I thought. Offer some useful information for the bereaved, maybe.
When I asked ChatGPT to ‘add links to some helpful websites’ I was pleasantly surprised with the links that popped up for the Bereavement Advice Centre, Samaritans, The Good Grief Trust and more.
Links haven’t always been a thing on ChatGPT, but they are now. But when prompts are flowing and copy is generating, it can be easy to slip into using the generative AI as a search engine. When I asked GPT 3.5: ‘Who is the ruling monarch of England?’ I got the response:
As of my last update in September 2021, the ruling monarch of the United Kingdom is Queen Elizabeth II.
ChatGPT did suggest checking the most recent and reliable sources to get current information and newer and paid-for versions of the tool may be more up to date, but it’s important to remember that it has its limitations and misinformation could creep into your copy. Although it has the power to gather information, OpenAI want us to use ChatGPT as a ‘reasoning engine’ rather than a fact database.
Don’t change your home page from Google just yet.
In conclusion
So, does ChatGPT signal the death of copywriting? It can’t replace creativity, emotion, strategic thinking (a-la SEO strategy), cultural nuances, specialised knowledge and our ability to adapt to shifts in trends… it told me so. But as Mr Altman warns us, we have to be careful with it.
“People should be happy that we are a little bit scared,” he’s said. It’s a tool that’s very much in human control and we should be concerned about which humans input that control.
And what did I learn about ChatGPT? Caution is king and copywriters still rule, there’s a mixed metaphor that I doubt AI would be stupid enough to generate.
But that doesn’t mean ChatGPT should be disregarded, in fact do so at your peril. Generative AI is a useful addition to every copywriter’s toolbelt and it’s perfect for:
- generating ideas for headlines, slogans and more when you’re suffering from writers’ block,
- suggesting alternative phrasing when you’ve seen a sentence one too many times,
- developing customer personas by providing information on demographics,
- and on a strategic level, analysing data related to your content, think web traffic, engagement metrics and social media feedback.
Copywriting’s far from dead and with powerful tools like ChatGPT at our disposal we’ve never been so lucky, but remember, with great power comes great responsiblity.
Further reading:
Sean Coyte, Senior Copywriter
