Chat GPT: What does it mean for a B2B content writer?

Greasy drive-through burger or Michelin-starred fine dining experience? That’s the analogy some content writers have made about ChatGPT vs themselves. Is there room for both? Almost definitely, but it’s a question we’ve been pondering at Marketing Fusion for a while now.

The written word plays a huge part in most of our services, from strategic, field and channel marketing, buyer enablement and customer advocacy right the way through to digital humans and the metaverse.

Is ChatGPT going to be a friend or foe for us and, most importantly, our clients?

Devil in the detail

B2B technology is a fast-moving, complex area, where the smallest details can make the biggest impact. It differs from other kinds of marketing in a few ways, including:

• It targets audiences more precisely, right down to the different requirements of different departments within the same business.
• The B2B sales cycle is longer than B2C products and nurturing relationships can be more important than quick conversions.
• Products and services tend to be expensive and complex so we must communicate their value in a convincing and highly resonant way.

Whether it’s for a blog, social media post, video, website copy or something else, the devil is in the detail. Around that detail, we build convincing stories, business cases and thought leadership. To do this, we talk to the experts within our clients’ organisations and research reputable and verified statistics, opinions and information.

ChatGPT-4 can tell me quite a bit about one of our major clients but when pressed on their upcoming product releases or any new features to existing products, it draws a blank and directs me to their website. Much of the information we need comes from the client and ChatGPT doesn’t have a direct line to them like we do.

It doesn’t deliver analysis or opinion, create beautiful infographics or insightful eBooks – it just very cleverly amalgamates what’s been said before.

Nonsense or narrative?

When we first started discussing AI language models, a colleague asked ChatGPT to write an article about noodles (we love a lunchtime noodle at Marketing Fusion!). I admit to feeling a little smug when I saw the word “noodles” appear as the first word in at least three sentences very early on in the article, which would never get past our proof-reading processes. ChatGPT’s article wasn’t the finished article, you could say.

What was produced wasn’t unreadable or incorrect, but it was repetitive and dry. Would you be compelled to read to the end? Probably not.

With last week’s Book Club in mind, when I asked it to “Tell me a love story about lost love” the resulting text began with a cringey “Once upon a time”, involved childhood sweethearts, a war, a painful letter and the planting of a tree. It’s not going to win the Booker Prize anytime soon.

That’s the trick that ChatGPT might not yet have learnt when it comes to writing marketing copy. Connecting ideas and telling a story, forming a narrative arc and taking the reader on a journey is crucial not just in literary fiction but also in the B2B technology world.

Every piece of content needs to give customers and prospects somewhere to go next to improve their business. It might one day be a robot writing the article but it’s still a human who is going to read it and make a business decision based on that content.

Power of the prompt

My question was weak though and that’s the point Christopher Penn, Chief Data Scientist at Trust Insights, made in his keynote about large language models at the recent Martechopia event in London. As he says, if you put a bland question in, you’ll get a bland response back.

Christopher quotes John Rupert Frith, who in 1957 said “You shall know a word by the company it keeps.” These large language models work on probability and so it’s important to ensure that the words you use in prompts make the connections you want them to make.

By feeding more nuanced, detailed information into ChatGPT you give the model more words “to grab onto,” as he puts it. Christopher uses the example of the prompt “Write a short paragraph about B2B marketing strategy with an emphasis on email marketing and lead generation” and then one with this additional text: “Focus on reduction of churn and increased audience loyalty. Include details about marketing automation and lead scoring. Write in a warm professional tone of voice. Write at 12th grade readability level.” The quality of the results is markedly different.

Start me up

Every writer has their methods, processes and routines when it comes to putting words on a page. Staring at a blank wall, eating endless snacks and playing loud, wordless music all spring to mind. Some writers prefer the research phase, some the frantic first draft, others the polishing and honing at the end. All of it can be painful or enjoyable, depending on the topic or mood.

I work from creative briefs that stipulate that I hit a certain tone, use a certain style and drive home specific messages. As a team, we brainstorm ideas and fine-tune every piece of work, often stripping back ideas and rebuilding them multiple times – all to a deadline of course – to ensure we hit exactly the right note for the client and their customers and prospects.

By offering ChatGPT a carefully worded question about a topic I get some key points and information that form the basis of a structure. That structure, and most certainly the content itself, will change beyond recognition. But I’ve started, which for the writer is often half the battle.

In the work we do for some of our major clients, we talk about technology and automation freeing people up to do more valuable, satisfying work, rather than replacing them.

This brings us back to Christopher Penn’s presentation, in which he suggests that the people who will survive in their writing roles are those that can already use language skillfully. ChatGPT can do so much but is only as good as the words that are inputted and the editing at the other end.

When it comes to human writer vs ChatGPT, the answer, at least for now, is not an either/or, it’s an “and.”