Don’t outsource your voice to a chatbot

People often tell me that writing is a great place to swap in AI. They usually say it with a tinge of concern for me and sometimes even ask outright whether I’m worried about losing my job, given that I write for a living. But content that’s generated by AI lacks the richness and originality that a human can bring in my view; you can’t simply substitute AI and get the same result.

While I’m lucky to be part of Cliniko—a company that deeply values the human touch—it’s true that many businesses have taken to outsourcing their written tasks to Large Language Models (LLMs) in recent years. When ChatGPT and Claude can spit out a reasonably persuasive paragraph in a few seconds, why bother doing it yourself or bringing a writer onto your team? Just feed in a few prompts into a chatbot and boom, your next newsletter or social media posts are all sorted. But there’s a significant cost to this. You might be saving time by using AI instead, but what else are you losing along the way?

If you work in healthcare, AI is always going to be an area where you should tread very cautiously. There are still big security concerns around its usage and it definitely isn’t a place to input patient data. However, even using it for your marketing or social media has some clear downsides. It’s also important to remember your written communications are how your patients connect with you between their appointments. Even if writing isn’t your strong suit and you dread this part of running a business, there are compelling reasons why you might want to reconsider delegating to AI—or at least making sure that you find a way to let your individuality come through if you do.

Why AI isn’t a substitute for your own voice

Your voice, your story, and your uniqueness are all extremely influential factors in how your patients perceive you. This combination is part of why they come to see you, rather than one of your competitors or the clinic down the road. Even if you’re the best practitioner around, people’s opinions of you are shaped by more than just their experiences in the treatment room. The updates you post to your social media, email marketing, or automated check-ins are how you stay in people’s minds between their appointments—and affect whether and when they return. Writing is central to all of these, which is why it’s so important to get right.

Writing is fundamentally a way that people relate to each other—the act of writing creates a relationship between writer and reader—and this is something that cannot be substituted by a machine. AI is also fundamentally generic in how it sounds; it flattens and homogenises, which isn’t a good strategy to stand out to your patients. You pop in a prompt, and the chatbot responds by cobbling together material from its dataset. The end result is neither unique nor truly tailored around you, even though it might feel like it. The sentences it produces can feel polished, but they’re also bland and don’t have any of your idiosyncrasies.

The AI tone

The writing that LLMs produce has a noticeable tone, which is usually quite recognisable. Even so, there are many misconceptions about what AI-generated text looks like and people tend to focus on the wrong things.

For example, the poor old em dash (the formal name for the long dash “—”) gets a particularly bad rap these days, with everyone assuming it’s one of the key tells of AI, even though it isn’t necessarily. The em dash has been around for centuries and was a favourite of many classic authors from the past. Pick up any book of literature from your shelf and thumb through it, and chances are you’ll come across em dashes lurking in its pages. You’ll find it across the works of Virginia Woolf, Toni Morrison, Edgar Allan Poe, F. Scott Fitzgerald, Charles Dickens, George Orwell, and James Joyce to name just a few. In fact, Jane Austen used them so often that her em dash usage has even been the subject of extended academic research. The reason that LLMs regurgitate this punctuation mark with wild abandon now is that these canonical literary texts have formed part of their training data.

The actual hallmarks of AI writing tend to be in its style and written expression. In terms of sentence structure, AI tends to use short and punchy phrasing that often comes across as quite formulaic. This is coupled with vocabulary that’s generic and broadly comprehensible, though there’s often heavy usage of italics or bolded segments to create emphasis and reliance on negative parallelism (i.e. “it’s not X, but Y”) or simplistic contrasts. Tone-wise, LLMs are often overly polite, hedging, or even obsequious, though they can also be prone to overexplaining and labouring a point. If you do use AI, look for these formulations next time you engage with it.

What does this look like in practice?

I asked ChatGPT to give me a couple of marketing sentences to promote Cliniko, written in a casual and friendly style. Here’s what it generated in response:

• Cliniko works tirelessly behind the scenes, asking for nothing in return. Not even a slice of cake. (Although if software could eat cake, we'd make an exception.)
• Running a practice is hard enough without playing hide-and-seek with appointments or wondering where that invoice wandered off to. Cliniko keeps everything where it belongs, like an exceptionally tidy magpie.
• Cliniko is the practice management software equivalent of finding money in an old jacket pocket. Unexpectedly delightful, surprisingly useful, and somehow always there when you need it.
• Cliniko won't fold your laundry, walk your dog, or remember your mum's birthday. But it will make running your practice delightfully uncomplicated, which feels like a pretty fair trade.

Even here, you can see some of the typical hallmarks of AI emerging. There’s the use of italics, as well as generic adjectives and adverbs (“delightful”, “tirelessly”, and “surprisingly” etc). The LLM relies on metaphors to simplify the message (e.g. comparing Cliniko to a magpie) and you can see the model’s attempts at humour across the examples, even if they kind of miss the mark (the cake joke falls flatter than a bad sponge). There’s also an example of negative parallelism in the final dot point—though that one actually contains an error; Cliniko can, in fact, remember your mother’s birthday if you add it to the calendar! Most crucially, however, none of these examples actually say much about what Cliniko does or give concrete detail about how it can help an allied health practice.

As a point of comparison, here are two examples of what we’ve used in our marketing in the past that we’ve written in house:

• Cliniko makes running your practice quicker and easier. With online booking and scheduling, treatment notes, invoices, and customisable body charts all in one place, Cliniko takes care of your admin so you can focus on treating your patients.
• Hit the snooze button in the morning because the admin is already sorted. Grab a coffee between patients because your notes are done. Leave the practice in time to make your dinner plans. This is what life looks like when you use Cliniko.

These differ in tone—the second one is more casual than the first—but both offer specific examples of what Cliniko can offer. The phrasing is simple and to the point, without overcomplicated matters by bringing in strange metaphors or humour for the sake of it. And, fundamentally, these are authentic to our brand and our tone, whereas the AI generated ones are not.

How to keep your clinic communications sounding human

I penned some thoughts a while back on what constitutes good writing (along with some tips). Although this was before LLM usage became quite so widespread, the principles of what kind of writing will capture your audience remain the same:

• Let your voice and personality come through
  Show some humour, personality, or share an anecdote. It sounds like a cliché, but being authentically yourself goes a long way. Even if it isn’t perfect, readers will respond to writing that feels human.
• Incorporate storytelling
  Humans are innately wired for stories, so anytime you can tap into this, it’ll likely resonate with your audience. Any time you’re writing, keep in mind what story you’re telling and why.
• Be mindful of tone
  Remember that it’s okay to be conversational. Your tone doesn’t have to be overly casual, but writing that uses overly formal language can come across as cold, rather than professional.
• Streamline your writing for your audience
  How you would communicate with your patients is different to what you’d say to your peers. Always keep your specific audience in mind when you’re writing.
• Centre the reader where possible
  Writing with empathy and putting yourself in your reader’s shoes is likely to lead to a better end result. One of the ways you can achieve this is by using the second person (“you”), which immediately makes your writing more relatable to your audience.

I’d strongly recommend that you keep writing your own words wherever possible. Ultimately, it’s better to be human—imperfect, and not always polished or slick—than to sound robotic. So you can afford to have some confidence in yourself here.