The human edit
Back in 2023, I wrote an article about AI and copywriting. At the time, ChatGPT was starting to be more widely used and I realised that, if I wanted to keep up with the times, I needed to get on board.
Like many writers, I approached it with caution. I was worried this new technology was going to put me out of work and, to be honest, it probably has lost me some work.
When I first went freelance in 2020, I was regularly commissioned to write blogs for clinic websites. These days, I am rarely asked to write them. That work has fallen off a cliff, and I suspect it is because many businesses now use AI to generate their own.
I understand the attraction. You can tick “write blog” off your to-do list in minutes and save yourself a few quid along the way.
However, unless you understand how to use AI well and, crucially, what makes good copy, you can end up with generic content that bears little resemblance to you or your business.
And here’s the thing: people can tell.
That does not mean I am anti-AI. Far from it. Over the past three years, it has become an invaluable tool in my work. The more I use it, however, the more convinced I become that it is only as good as the person using it. The trick is knowing what to ask it to do, what not to trust it with and where human input still matters.
Would you like a side of sameness with that?
As an editor, I receive a lot of copy that appears to have been generated using AI and submitted with little or no human intervention. In fact, sometimes they haven’t even bothered to remove the AI response from the copy! It may be grammatically correct and neatly structured, but there is often something missing. It may be bland, repetitive and devoid of personality.
Then there are familiar tells: the long em dashes – thanks very much – and the relentless lists and bullet points and single sentence paragraphs. Then there’s what aesthetic consultant Vanessa Bird calls the “M&S phrasing”. It’s not merely a consultation; it’s a journey.” “It’s not just skincare; it’s a ritual.” “It’s not simply a treatment; it’s a transformative experience.” I am asleep already.
And then there is my least favourite opening: “In the ever-evolving world of aesthetics…”
Once you start noticing these things, you can’t unsee them.
That doesn’t mean you need to stop using AI – that ship has sailed – but I would suggest stopping accepting its first answer as the finished product.
Start with your own raw material
The quality of the output depends heavily on the information you give it. A prompt asking ChatGPT to “write a blog about skin ageing” is likely to produce the same generic information it would give almost anyone else.
Before using AI, decide what you actually want to say. Give it your observations, opinions, examples and experiences. Tell it who the audience is, where the copy will appear and what you want the reader to take away from it.
If you are writing about consultations, for example, tell it what you have noticed in your own clinic. What questions do patients repeatedly ask? What do they misunderstand? Has a particular case changed the way you practise? What do you do differently from other practitioners?
That is the material that will make the finished article yours. AI can help organise it, but it cannot invent genuine professional experience on your behalf.
Give it a job, not the whole project
I find AI most useful when I give it a defined task rather than asking it to take over the entire piece.
I use it to plan my workload, develop article briefs and prepare interview questions. It can help turn transcripts into usable material, identify repetition, explore a different structure or reduce a 2,000-word article to 1,000 words without losing its central argument. Everyone who knows me knows I overwrite, so that last one is particularly useful!
Try asking AI to suggest a structure for your article, identify questions a patient might have about a subject or review whether the tone is appropriate for the intended audience. These are all tasks that can support your thinking without handing over the whole process.
Train it to recognise your voice
One of the things that is making copy boring is that it all sounds the same. If you want copy to sound like you, you need to give AI examples of your existing writing and explain what you like about them. Be specific about the tone you want, but also tell it what you do not want.
You might ask for copy that is conversational, knowledgeable and straightforward, while avoiding jargon, exaggerated claims and overly promotional language. If there are phrases or habits you dislike, say so.
You can also challenge the result. Tell it that the first version is too generic, formal, wordy or polished. Ask where it has repeated an idea or used language that could have come from any clinic.
Thought leadership still requires thought
This is particularly important when submitting articles to publications.
Editors commission contributors because we want different experiences, opinions and areas of expertise. We want to know what you have observed in your clinic, how you approach a problem and what you have learned through your work.
AI can help organise your thoughts or make your writing clearer, particularly if writing does not come naturally to you. There is nothing wrong with that. The problem arises when it is used as a substitute for having an original thought in the first place.
Hallucinations and the ghost reference
One of the strangest problems I have encountered when writing clinical articles is the “ghost reference”. These are citations that look perfectly plausible, but either do not exist or do not say what AI claims they say. The title sounds convincing, the journal is real and the author may even work in the relevant field, but the paper is nowhere to be found. Even when a reference is genuine, you still need to open it, read it and check that it supports the claim being made. Do not ask AI to produce a reference list and assume the job is done.
Clinical articles with extensive referencing can still take me between 12 and 16 hours, even with AI assistance. The technology helps me work through the material more efficiently, but it does not remove the need for due diligence.
Practitioners should also think carefully about what information they put into an AI platform. Do not enter identifiable patient details, confidential business information or unpublished material without understanding how the particular tool stores and uses that data and whether its use complies with your organisation’s policies.
The final word
If you are using AI to generate copy for your clinic, whether that’s blogs, marketing materials, social captions or thought leadership pieces for publications, you can’t just click and go. Well, you can, but ultimately, it could start driving audiences away.
Before pressing send on an article submission or posting content on your website and socials, read it out loud and see if it actually sounds like you; be honest about whether it contains your own thoughts, opinions or insights and ask yourself if another practitioner could have churned out exactly the same content by typing in the same prompt. This will help you identify if you need to spend a bit more time and effort adding some more “you-ness” into it. Believe me, it will help.
To sum up, I asked AI what its advice was, and this is what it said: “To use AI effectively, you need to give it context, question its answers and be prepared to reject what does not work. You need to understand your audience and know what you want to say. Most importantly, you need to put something of yourself into the result.”
I couldn’t have said it better myself.