Will ChatGPT really replace copywriters?

Why you shouldn’t let AI write your website (just yet)

Author
Gabriela Freund | © punkt & komma
Gabriela Freund
Content editor EN

ChatGPT is all the rage right now – or in marketing speak: an absolute buzzword. Of course, at punkt & komma, we too are well aware of these developments. We are looking into the topic and technology, are experimenting with it extensively and regularly exchange ideas.

After all, opposing technical innovations goes against our professional honour in digital marketing. If a tool allows us to make processes more efficient, better and less complicated – bring it on! However, we also see the chatbot with a critical eye and are in the process of finding out to what extent AI can be useful for us in content marketing.

Which brings us straight to the point ...

AI is a tool and not a takeover

People keep asking us: “So? What do you think of ChatGPT? Aren’t you at all worried that you’ll soon be out of your jobs?” The answer is: No. And we’ll tell you why. We see Artificial Intelligence in a similar way as the Head of Marketing at Jasper, one of the best-known platforms for AI text and image generation:

It’s a tool, not a takeover.” – Austin Distel, Head of Marketing at Jasper.ai

And as is usually the case with tools, it is the users and how they apply this tool that makes all the difference. AI can be useful and has its place – but to create really good content and convincing copy, you need someone with know-how. Someone who knows exactly what to look out for and what to do with the generated content.

After all, the spell-check in Microsoft Word is no substitute for professional editing. That’s why we at punkt & komma continue to work closely with external editors and also have our texts proofread according to the six-eyes principle. But that’s just as a side note.

Where do we see the weak points of ChatGPT?

Before we turn to the question of where we currently see the potential of AI, let’s take a quick look at the current shortcomings:

  • Lack of uniqueness of the content
    If you look at how ChatGPT gets its information, it is pretty obvious: The content is not really unique. Even if the texts are not necessarily plagiarised – the content has to come from somewhere. So the bot fishes in a pool of existing ideas and then creates new sentences and texts from that. Groundbreaking new thoughts and approaches? Not a chance! In other words: The bot doesn’t think outside his bot-box. 
  • Lack of topicality in the case of ChatGPT
    ChatGPT has “limited knowledge of world [sic] and events after 2021” (Source: OpenAI). Simply put, the system draws from a corpus of information that may no longer be up to date. So if you want to keep your finger on the pulse of time, you have to get your information elsewhere. 
  • Misinformation
    ChatGPT sometimes produces content that is factually incorrect or makes no sense in context. This can lead to embarrassing mistakes and damage a brand’s reputation. Even worse, in the case of companies such as pharmaceutical companies, false information can even result in legal consequences. So beware! 
  • High input & training requirements
    Do you want ChatGPT to produce relatively good content that fulfils user intentions, provides all the important information and does so in an appealing format? Then you have to feed the bot with a detailed briefing or a complex prompt. For this to work, you need to know what you are doing – and it takes time. The content only becomes truly usable if you train the system over a longer period of time. 
  • Strong need for revision
    This may be a subjective assessment but, at this point in time, the content experts at punkt & komma aren’t all that impressed with ChatGPT. Several writers on the team have already experimented with the tool – so far, it hasn’t produced anything that we wanted to use as is and where we didn’t feel the need to heavily edit the output. Opinions ranged from “It’s okay as a rough first draft” to “I would have been faster if I had written the text myself”.

As you can see: Not all that glitters is gold – and perhaps thinking on your feet does require having actual feet, after all. But if you want to convince yourself of the copywriting skills of ChatGPT: We have already tested it for you in a blog post!

Open questions about AI and content

As we all know, publicly available AI is still in its infancy. This raises a number of questions that cannot be answered satisfactorily yet. We have briefly summarised the most important ones that come to mind:

  • What about the copyright of the texts?
    OpenAI, the makers of ChatGPT, state in their FAQs that users “own the output [they] create with ChatGPT, including the right to reprint, sell, and merchandise”. Nevertheless, various questions regarding copyright have not yet been clearly answered. The AI produces its content on the basis of data sets. These data were originally created by humans, even if the AI eventually combines them into new sentences.
    The Austrian Copyright Act states that “works within the meaning of this Act [are] original intellectual creations” and “the author of a work is the person who created it”. So: In this case, is it ChatGPT or OpenAI? Is it you as a user? Or is it perhaps the persons/entities based on whose data the text was created? We are still in a legal grey area here. 
  • Does Google penalise AI-generated texts?
    As things stand, the quality of the content is still the most important criterion for a good Google ranking. Google doesn’t care how the text was created, as long as it provides users with the information they were looking for – and does so in the best possible way.
    Based on this information, a really good AI-supported (and we are deliberately saying “supported” and not “generated”) text has the potential to rank better than a poorly written text by a human writer. Whether this will change in the future cannot yet be predicted, but since Google itself has already launched the beta version of its own AI (“Bard”), it can be assumed that the search engine is unlikely to penalise AI texts per se.
  • What happens if the pool from which the AI retrieves its language patterns and information eventually consists almost exclusively of AI-generated texts?
    Currently, AI tools still get their input from a corpus of human-generated texts and information. In a few years, things might look different. This could potentially lead to a downward spiral in terms of the quality of the texts and the information that’s circulating. Let’s hope that in this case, Google’s rule no. 1 will apply once again: Poor quality content and information will be rigorously penalised.
  • How will Google Search change in the future?
    Google’s AI “Bard” is very much geared towards use in Google Search. In the future, users will most likely first see an AI-generated answer – similar to a featured snippet (Source: Google) – before the rest of the results.
  • What about sustainability and scalability?
    What is often ignored are the enormous computing capacities that AI requires. It becomes apparent in the frequent downtime of ChatGPT. This naturally makes you wonder about the sustainability and scalability of AI on a large scale. Solutions will certainly be found for this issue, but what they will look like and how satisfactory they will be remains to be seen.

Where do we currently see the potential of AI?

Now that we have taken a closer look at the weak points and open questions regarding AI and ChatGPT, in which areas can these tools actually be of use to us?

  • Generating ideas and combating writer’s block
    A blinking cursor = every copywriter’s nightmare. It’s an all too familiar feeling: Sometimes you just lack that decisive spark of an idea that helps you get started on a new text. In a situation like this, ChatGPT may be able to help. The same applies when it comes to incorporating additional perspectives and finding alternative approaches. The bot is also already quite good at structuring or summarising large amounts of text.
  • Inspiration for newsletter subject lines, headlines and social media captions
    As a brainstorming buddy, ChatGPT can not only support you in generating ideas for longer texts. When writing headlines, newsletter subject lines or social media captions, it can also be helpful to have the bot generate a few suggestions for you. Even if many of them are useless, there may be one or two gems among them. With a little creativity and know-how on the part of the copywriter, micro-content can be created from them that captivates and converts. 
  • Better briefing materials by clients
    Sometimes it’s not enough for a copywriter to just receive a few keywords and bullet points without context. Complex topics in particular often require detailed explanations – and that takes time on the client’s side. As we all know, time is usually a scarce resource in companies. ChatGPT can help by turning bullet points into texts that (after being checked for factual correctness) can serve as a more in-depth briefing for copywriters.

Our (preliminary) verdict

Even at the risk of sounding like a broken record, we have one motto that we have been firmly abiding by from day one: It’s not about more content, it’s about better content! Quality over quantity. Our credo is that only really good content that is useful, captivating and inspiring will ultimately impress both users and Google. In our opinion, this cannot currently be achieved with 100%-AI-generated texts.

To win the battle for the attention of users and search engines, you need more than that – more marketing knowledge, more expertise, more talent for writing, more wit and a heaping scoop of humanity and authenticity.

But before we get too philosophical here: Would you rather leave the writing of your texts to experienced content marketers? Get in touch with our very human – and rumour has it, very likeable – content marketing experts!

Recommended articles