Is ChatGPT a better writer than me?

Cayley Jones, UX writer at Helm

With their computational power and access to vast amounts of data, products powered by large language models (LLMs) like ChatGPT, arguably pose the biggest threat to me and my fellow writers. While it’s undeniable that an LLM would outperform me considerably in technical tasks like writing an 800-word essay in 10 seconds or generating 20 campaign slogans in the blink of an eye, there is a surprising amount of areas in which us human writers still have the upper hand… for now.

If my goal is to remain one step ahead, it’s vital for me to know what I’m up against – this means keeping tabs on AI’s ever-expanding capabilities. Going forward, the key to staying relevant as a UX writer will lie in understanding AI, leveraging its full capabilities and stepping in where it falls short.

To generate, or to create…

While AI can mimic human-like responses to some extent, LLMs do not possess the capacity for independent, abstract thought, unless specifically prompted to do so (by a human). Where AI relies solely on existing data and prompts, I can draw on my physical and sensory experiences to enrich my understanding and infuse my writing with complex emotional nuances. This exploration of my own experiences forms part of my creative process, which often leads to unexpected and, if I’m lucky, brilliant results. AI logic cannot necessarily understand or be trained to imitate this intuitive, instinctive way of thinking that many creatives possess (for the time being, at least).

While LLMs can generate original text based on specific inputs and the data on which they have been trained, that’s where their “creating” power stops. I, on the other hand, as a living, breathing, thinking human, have the inherent ability to create and craft unique ideas and concepts from scratch, without being prompted to do so.

Generative AI models can emulate many human-like responses, but in the absence of a well-crafted, detailed prompt, their responses are pretty underwhelming. For example, if I was asked to “tell [you] about the ocean…” I could write endless pages describing this body of water in vivid detail. But when I typed this exact question into ChatGPT 3.5, it responded with:

“This prompt is quite broad and lacks specificity, which could lead to a general or surface-level response. However, I'll do my best to provide a brief overview…”

And it then proceeded to generate a few bland and very (pardon the pun) surface-level sentences for me.

LLMs can perform countless natural language processing tasks at highly impressive speeds, but they cannot write their own prompts. And the quality of an AI model’s response will only be as good as the prompt you provide it with.

CX is a human experience

The fact that I am a real person, writing for real people, remains one of my biggest advantages. When brainstorming ways to craft the best possible UX solutions that elevate the experience for end users, I can put myself in the user’s shoes and ask important questions like: What do I like? What don’t I like? What do I usually look for, or find lacking, in similar services or products? These experience-led insights are invaluable within the world of UX and give me a distinct edge over AI models.

My tangible existence also gives me the advantage of being able to understand and navigate contextual complexities in a way that machines can’t always do reliably or consistently.

On its own, an LLM possesses no inherent contextual awareness; or rather, its contextual understanding is driven solely by (and limited to) that of the person prompting it at the time.

The contextual “comprehension” within AI-generated content cannot match the depth of empathy and awareness inherent in the work of a human writer, especially within South Africa. We can’t program a machine to understand what it means to be a South African. But as someone born and bred in this rainbow nation, navigating its cultural, socio-economic and linguistic complexities is part of my DNA. And by truly understanding who my users are and what makes them tick, I can create content that resonates with the South African on the other side of the screen, at a level beyond AI’s reach.

Can I compare?

Granted, LLMs may process information and generate text faster than I’ll ever be able to, but I’d like to believe that my critical thinking and reflection skills still surpass those of generative AI (for now!). AI models operate within predefined parameters, and these are still determined and defined by people, like me. That’s not to say that my ‘human’ way of thinking is better than an LLMs data-driven logic, but when it comes to UX writing – a task that requires empathy, creativity and intuition – I do believe I have the edge.

So, do the rapid advances of generative AI threaten my role as a writer? Well, given its current limitations, I believe my job is still safe. But I am also very aware that while these shortcomings may be true as I write this, the same may not apply in a year, or even a week, from now.

There’s no denying that AI is here to stay. And ultimately, the question “Can AI replace me as a writer?” is one I’ll have to ask myself every step of the way, while doing everything in my power to make sure my answer remains “no.”