I’ll just come out and say it, large language models (LLMs) like ChatGPT, Claude, Gemini and others currently suck at creating high quality written content and they’re not part of our process for creating content for our clients. These models are great for some tasks, such as coding and image generation, but they’re still not great writers.
While we’ve never used AI to generate work for clients, we have experimented with several of these models to see if they could help optimise our creative process or improve the quality of our writing and we’ve found they can’t reliably do either.
In our experiments, using AI to generate content has actually wound up taking more time rather than less, partly because we’ve had to double-check everything the AI outputs. We’ve found that AIs often make crucial mistakes when outputting content or just straight hallucinate and make things up. All these mistakes need to be identified and corrected, a surprisingly arduous and time consuming task..
The other huge issue using AI for writing is that it writes in a weird and unengaging way. Even when we’ve trained AI models on our writing styles the output has been bland and kind of faux-bubbly — it has a frivolous yet boring tone which is often completely inappropriate for the content it’s creating.
AIs also tend to have distinctive language ‘tics’ such as unusually frequent use of em dashes, an over-reliance on certain words such as ‘delve’ and a strong tendency to write sentences in the format ‘It’s not X, it’s Y,’ a format many readers have now come to loathe.
Fixing the style issues present in AI writing is even more time consuming than fixing the factual errors because they pervade the entirety of the content.
The truth is it’s easier, quicker and leads to better results when we write content ourselves.
Of course, we’re not the only ones who’ve noticed these problems. It’s been widely noted that AI models tend to have these particular issues, but that hasn’t stopped the widespread use of these AI models to generate written work.
For example, according to a December 2025 New York Times article many AI models have an uncanny tendency to choose the name ‘Elara Voss’ as the protagonist if you ask them to generate a science fiction story. Coincidentally, there are now hundreds of self-published sci-fi books on Amazon featuring a protagonist named Elara Voss or Elena Voss, prior to 2023 there were none.
I don’t know about you, but I’m personally not that interested in exploring the Elara Voss extended literary universe, largely because it was created by AI and I therefore have the not unreasonable suspicion that the readerly pleasures awaiting me may be a little bland and samey.
But it’s not just cynical sci-fi authors turning to AI to create their content. In December of 2025 three South Australian lawyers — a solicitor and two barristers, one of whom is a King’s Counsel — were referred to legal regulators after they submitted evidence which was sourced from AI and contained numerous hallucinations.
Scientific research, another field where accuracy and truth are vital, is facing its own AI induced hallucination crisis, with a recent study which examined 2.5 million studies published in 2025 finding almost 150,000 papers contained hallucinated citations.
Of course, it very often matters if written content is accurate and truthful and this is a huge part of why we don’t yet use AI in our work for clients.
Beyond the issues with the AI itself, there are also issues with the ways audiences are now reacting to AI generated content. These is something of a backlash against AI generated content taking place. Often the fact that a piece of content has obviously been created by an AI enough to put a reader off — this is perhaps most notable with AI generated videos and images, but it’s also now impacting audiences reactions to written content.
If the writing wasn’t important enough for a human to sit down and write, why should someone think it’s important enough to sit down and read? Many would-be readers are deciding AI content isn’t worth paying attention to.
Despite its growing use by professional writers, as of 2026 we at Quill Studios don’t use artificial intelligence to generate any of our content for clients, we think it simply isn’t up to the job yet. We’re concerned about accuracy, style and audience reactions to AI generated content, and we think you should be too.
If there comes a time when AI can augment our process of creating content we’ll gladly use it to the extent it’s useful. For now though our work will continue to be cooked up by our human brains and tapped out by our human fingers, hopefully to be read and enjoyed by other humans.