Don’t worry, a human wrote this

The landscape of writing today is fraught with the danger that you may end up sounding like AI. Perfect, wordy, quietly certain. But here’s the kicker—it’s not just a function of style, but of substance as well.

If the sentence above reads AI-generated to you, you’re bang on, although it was actually written by a human—albeit one imitating AI, which imitates certain human writing styles on which it was trained, which may dictate how we make up our sentences in the future, and on and on like a series of reflections in a science experiment.

As a professional writer, there are many pitfalls one must avoid. I have a tendency to unconsciously repeat words in the same sentence or adjoining sentences. My editor has often pointed out that my sentences can be overlong, sometimes the length of an entire paragraph. I tend to overuse em dashes and semicolons (because having learnt how to use them, it’s basically criminal not to). Some writers have ticks—they begin too many sentences with a chatty “yes”, or end every paragraph with a question.

To this list we must now add the danger of sounding like AI. As we are exposed to more and more content generated by AI—on our social media feeds, in our email inboxes, in advertising copy and summaries of films and TV shows—we are constantly and unconsciously absorbing the way it uses language.

As we learn to identify and avoid these, the already hard job of writing becomes harder as we try to make ourselves sound more human. Already, em dashes have come under scrutiny and are considered a red flag for AI-generated writing, along with several constructions like the ones in the first sentence of this piece.

Research led by Dr. Hiromu Yakura at the Max-Planck Institute for Human Development in Berlin found that words overused by ChatGPT—such as ‘delve’, ‘comprehend’, ‘boast’, ‘narrative’, and ‘meticulous’—saw a spike in usage in podcasts and YouTube talks by academics post-ChatGPT, says an article by Reuters Institute at the University of Oxford titled How AI-generated prose diverges from human writing and why it matters.

“(This suggests) that these words are now used more often by humans as a result of being exposed to AI-generated content in which they are overrepresented. These findings suggest a scenario where machines, originally trained on human data and subsequently exhibiting their own cultural traits, can, in turn, measurably reshape human culture,” says the article, essentially reporting a conversation between three experts from the fields of linguistics and journalism.

AI is also constantly learning. A search for “tell-tale signs of AI writing” threw up many suggestions collated by AI from various articles (hopefully) written by humans, but it’s only a matter of time before chatbots train themselves to avoid these constructions. It might be impossible for us to keep up at some point.

Afraid of sounding like AI, I now have to check myself constantly to see if my writing comes across as too glib or smooth. I wonder what I could do to undo this. Maybe throw in a wonky sentence? Add a word or two in an Indian language? Leave a thought unfinished? After all, imperfection is the hallmark of humanity today, though we must wonder in what ways incorporating imperfection into writing will change the ways in which we express ourselves.

I used to roll my eyes if I caught a typo or an odd sentence in a piece I was reading; now I find myself thinking affectionately “aww, a human wrote this”.