David Mummery is a GP in South London
What would you say if I said that the whole of this article was written by artificial intelligence (AI), such as ChatGPT?
Would you believe me?
Well, whether it has been or not (which I will leave to your judgement, reader), an increasing number of medical journal articles, academic papers and blogs are likely to have had some AI input, if not be completely written by AI. This seems now to be common and almost accepted practice, although in previous years and decades, this would have been called “cheating” or “fraud” if it has not been declared openly.
So, in the spirit of further enquiry, it may be helpful to outline a few clues or give-aways that may give you an indication as to whether a human being or a computer such as Chat GPT has written what you are reading in the medical literature, academic journals, or in a GP online magazine such as BJGP Life.
So, as an experiment I asked AI1 how to tell when something has been written by AI ; some of the answers are as follows:
I asked AI how to tell when something has been written by AI…
There are several clues that can indicate whether an article is AI-generated: repetitive phrasing, lack of emotional nuance and a tendency to overuse certain words and phrases can be indicative. AI-generated text may sound stiff and formal, the sentences overlong, there may be an overuse of jargon or buzzwords, there may be a lot of commas, or semi-colons; the sentences may lack the natural rhythm and pauses of human speech, and may compensate for lack of understanding by using generic or technical terms excessively. There may be inconsistencies in tone and style and a lack of personalisation, where AI-generated content may not be tailored to a specific audience or context.1
Are you with me so far?
This sounds like a quite lot of medical writing, right? A bit bland and monotonous?
So, we may need to look for more clues and to do more detective work…
If you ever have chats with Generation Z or Alpha, they tend to know a lot more about things like this than oldie Gen Xers like myself. Apparently, there are some other clues.
Have you heard of the “AI hyphen”? No, I hadn’t either.
I mean what human being do you know of that would use two types rather than just one type of dash?
The AI-hyphen refers to the perceived use and over-use of ‘em’ dashes. Em dashes are longer – about double the length – than the normal hyphen or ‘en’ dash, (-).
Some believe that the em dash is a telltale sign of AI – generated text due to its frequent use in machine—written content. However, this theory is not universally accepted. It is thought that some AI writing tools may utilise em dashes due to their training on various writing styles, including styles that frequently use em dashes, like magazine or blog writing.2
It is important to note that there is no definitive or foolproof way to identify AI—generated text, and the em dash is not a reliable way according to some. However, if you see a medical journal article or blogs with a mixture of en and em dashes, this might arouse suspicion of AI—writing.
I mean what human being do you know of that would use two types rather than just one type of dash? Not one I know, that is for sure, but maybe I’ve just had a sheltered upbringing.
You might be thinking, why is this important anyway? Who cares whether it is a human doctor or a computer algorithm that is writing the article?
Well, I suppose that is one of the defining questions of our time currently in the 21st century.
Does anyone care and does it matter?
References
- Google AI summary of “How to tell when something has been written by AI,” entered into the search box of www.google.co.uk [date accessed 05/06/2025]
- https://www.techradar.com/computing/artificial-intelligence/blade-runners-of-linkedin-are-hunting-for-replicants-one-em-dash-at-a-time [date accessed 05/06/2025]
Featured Photo by Randy Tarampi on Unsplash