In a recent webinar we hosted on writing in the age of AI, we conducted a quick poll. We presented two short pieces of writing, one taken from The New Yorker and one generated by ChatGPT about the same theme, and we asked our audience to select which text they thought was AI-generated.
They chose the text from The New Yorker.
(Wonder which one you’d have picked? Take a look at the examples here.)
The result wasn’t what we’d expected, requiring a quick scripting pivot for me and my colleague Alex. But it proved far more enlightening.
When we asked about the reasons for picking The New Yorker text, communicators pointed to the em dash, generic phrases and the “robotic” writing style – all hallmarks of AI-slop, as it is so often termed now. But it wasn’t AI-generated text they were attributing these characteristics to – it was arguably stylistic, journalistic writing.
Aside from thinking we’d picked a poor example, the immediate conclusion seemed to be that AI-generated writing is getting so good it is becoming harder to spot. Or simply that we have become too reliant on a narrow set of signals when evaluating copy.
But the more useful conclusion is that spotting whether text is AI-generated is probably a futile test and not the question we need to be asking.
AI has learned from us
If a piece of writing is clear, fluent and well structured, that no longer tells us much about who – or what – wrote it. AI can now produce plausible and competent text at speed. After all, it has been trained on human text, on the writing we have already produced.
That’s why spotting AI writing feels like a redundant benchmark. A better question is not whether a piece of writing sounds human or not, but whether it is doing the job it needs to do.
The real test of good writing
This means asking a different set of questions, all of which form the foundation of good writing. Who is this for? What do they need from it? What are we helping them understand? What do we want them to think, feel or do next? What evidence, expertise or experience sits behind the argument? Is it accurate? Is it useful? Does it have a clear point of view? Would it help someone understand, decide or act?
This is where communicators need to focus, and where they are equipped to excel. Just because AI has made writing easier to produce, it doesn’t mean it produces quality writing that achieves the intended aims. That requires a proper brief, clear intent, context and a rigorous review process.
The latter is particularly important because AI can be inaccurate. A communicator recently told me that an LLM generated quotes supposedly made by their CEO. If checks had not been carried out, inaccurate material could easily have been distributed.
From writers to editors
This brings us to the changing role of the communicator today. We are not so much writers as editors, chief storytellers and reviewers. And this quality control should go beyond spelling, grammar and house style. It is about whether your content has a clear purpose, a stated audience and a meaningful point that captures attention. If it doesn’t, it is unlikely to cut through in a content-saturated environment.
Ultimately, these are not questions AI can answer on its own. It doesn’t know your audience the way you do, nor does it understand organisational context, stakeholder nuance and reputational risk.
So perhaps the real lesson from our poll is that our audiences, like us, may not be able to discern what you have written yourself and what AI has helped create. But that doesn’t really matter. What they will know is whether the content grabs them, informs them or gives them something worth paying attention to.
Less about detection, more about intention
That should bring our focus back to the fundamentals of good writing. It means starting with the audience, not the output. Being clear on what they need, what you want them to understand, and what the content has to achieve before you ask AI – or anyone else – to draft it.
It means reviewing for substance, not just style. Is the argument clear? Is the evidence sound? Is the point of view specific? Does the piece help the reader think, decide or act?
Because good writing in the age of AI will not be defined by whether it sounds human. It will be defined by whether it is useful, credible and worth reading. AI can help produce the words. Communicators still have to make them matter.
For practical tips on creating clear, effective content in the AI age, download our tip sheet, with insights from the webinar and our new course, Writing and storytelling in the age of AI.
