Ever since the launch of ChatGPT, AI-generated texts have been seeping their way into more and more corners of our lives. Many platforms have introduced automated text generation features to try and make things easier on their users – from dating apps to event ticketing platforms and social media websites.
But have you ever noticed there’s something distinct about AI-generated texts? They seem to have their own style – a manner of writing which is at once generic and sanitised, but also somehow devoid of spark or life. It can be hard to put a finger on it, but sometimes it’s just clear when the text we’re reading has been written by a robot.
Below, we take a look at three examples of AI-generated texts to try and deconstruct what is quickly coming to be known as AI Speak.
The coffee morning
We used the AI auto-generate feature on an event ticketing platform to create a description for a charity coffee morning. We asked the AI to make the text punchy, fun and inviting. Here’s what it gave us:
“Get your caffeine fix and sprinkle some laughter on it too! Join us at the Local Community Centre for a morning of coffee, good vibes, and making the world a better place. We’ll be brewing up a storm of deliciousness at 24 High Street, UK. So bring your caffeine cravings and rally your crew—friends, family, and co-workers—for a feel-good fiesta that gives back. See you there, ready to sip, laugh, and do some serious good!”
Although the text is grammatically flawless and has a nice rhythm to it, the sense of enthusiasm feels massively over-done. Expressions like ‘caffeine cravings’ trigger our cringe reflexes, while ‘feel-good fiesta’ feels like an off-the-mark choice in the context. What’s more, the text feels very American in its style, ignoring the fact that the event is set to take place in the UK. Over all, the feeling conveyed is distant, false and unlikely to chime with the intended audience for the post.
The dating profile
We asked ChatGPT to write a brief bio for a dating profile for a man in his thirties who enjoys video games, cooking and hiking. We specified that the text should feel open and warm, but also humorous and witty:
“🎮 Level 32 Adventurous Cook 🍳
Hey, I’m [Name], a guy in his early 30s juggling a day job and a love for virtual quests, hiking trails, and whipping up culinary delights. When not in the gaming zone, I’m chasing sunsets or experimenting in the kitchen (warning: occasional burnt offerings, but I promise it’s part of the charm).
When I’m not saving the world on my console, you’ll find me hitting the trails, chasing sunsets, and occasionally getting lost in nature (don’t worry, I have a great sense of humor and a reliable GPS). Looking for someone who appreciates a well-timed gaming reference, enjoys outdoor escapades, and can handle my kitchen escapades. If you’re up for an adventure with a guy equally at home in a virtual world or on a hiking trail, swipe right!
Bonus points if you can beat me in Mario Kart—fair warning, I’ve got some serious skills. 🎮🏞️🍳”
Overlooking the excessive enthusiasm of the text, the first impression we get is that it reads more like ad copy. Rather than sounding like the voice of a real human-being, we feel as if we’re being addressed by a corporation with something to sell. Bearing in mind that generative AI programmes draw on datasets from multiple textual sources, we wonder if it sometimes has trouble distinguishing between these different styles of speech.
Drilling down a little deeper into the substance of the text, the ‘well-timed gaming reference’ remark feels like showing rather than telling, while the repetition of ‘escapades’ and ‘when not’ simply smacks of bad writing. Even if the text may seem initially impressive for having been generated out of thin air in mere seconds, we have to ask ourselves; what is its value, really? Would it entice you to swipe right, or would it simply disappear into the background like meaningless white noise?
The Twitter post
We told ChatGPT that we are a young professional in the languages industry, and we asked it to write us a clever and insightful tweet that we can post to impress our colleagues and peers:
“Diving into the intricate tapestry of language every day, where each word is a brushstroke, and every translation is a masterpiece. In the world of languages, I’m not just a professional; I’m a linguistic virtuoso crafting connections across cultures. #LanguageMagic ✨🌐 #Wordsmith”
The first thing that strikes us about this text is its excessively boastful nature and once more, we find its style more befitting of a company or a business rather than an individual. At the same time, there is no originality at the heart of the message. We asked the AI to come up with an insightful observation, but the words ring hollow and appear devoid of substance.
The problem
Perhaps the problem lurking at the heart of all three of these texts is also the one most obvious thing that defines them – they are all written by a machine, and it simply seems impossible for the AI to conceal this fact.
When we write as humans, we bring so much of ourselves to the page – our interests, our experiences, our sense of humour and our own unique personality, built up over a lifetime of interactions and exploration. Even when we’re instructed to write in a neutral and objective way, because we are all humans, and because language is our toolkit for making and understanding the world, whenever we write, we inevitably shine through in what we produce. This is something we are all used to and that we quite simply expect, whether we realise it or not.
What these three texts appear to show is that no matter the context or no matter what we ask for, AI just can’t seem to convince us that there is a human behind the output. This leaves the text feeling dead and hollow. It gives us that unsettling, uncanny-valley sensation where we recognise that the writing is technically passable – that the prompt has technically been executed correctly – and yet there is something missing. That something is the human touch, and it cannot be replaced.
