The Problem With AI Slop Isn't the AI. It's the Lack of Thought.
Updated 6 min read
AI slop is not about writing style: it exposes a lack of thought, evidence and original insight in AI-assisted content.
Contents
Contents · 4 sections
People are getting very good at spotting bad writing and blaming it on AI. But bad writing was around long before ChatGPT. And the real problem is something else entirely.
There's a growing obsession with identifying AI-generated writing.
Apparently, there are certain tell-tale signs. Too many em-dashes. Too many neat lists. Too many phrases like "it's not X, it's Y".
You know the sort of thing:
"Digital transformation isn't a technology problem. It's a people problem."
The implication is that these tired rhetorical devices are somehow a product of generative AI.
But anyone who has spent much time reading business articles over the last few decades will recognise them immediately.
They were everywhere long before ChatGPT came along.
In fact, the older version was often worse:
"Everybody thinks digital transformation is a technology problem, when it's really a people problem."
Really? Who are these people?
I've worked in business strategy and transformation for decades. I can't remember the last time I met someone who genuinely believed that technology alone was enough to deliver a successful transformation.
The statement presents an almost universally accepted idea as though it were a startling revelation.
And that's what makes it so irritating.
Not the rhetorical construct. The complete absence of insight.
It's not the writing style. It's the lack of substance.
(See what I did there?)
There's nothing inherently wrong with saying "it's not X, it's Y".
It's a perfectly legitimate way of drawing a distinction. Used well, it can challenge assumptions, sharpen an argument or help people see a familiar problem differently.
The same goes for em-dashes, rhetorical questions, short sentences and any number of other writing techniques that have somehow become associated with AI.
If you've come up with a genuinely interesting idea, I don't particularly care how you choose to express it.
What matters is whether you have something worth saying.
Consider the digital transformation example.
If you can explain why organisations repeatedly underestimate the human challenges of transformation, identify a previously overlooked cause, or propose a better way to address it, I'm interested.
But simply declaring that digital transformation is about people rather than technology adds almost nothing to the conversation.
It's a familiar observation dressed up as a profound insight.
And we've been doing that for years.
AI hasn't invented empty thinking. It's made it easier to publish.
This is where generative AI really has changed things.
Previously, producing a thousand-word article required a reasonable amount of effort, even if you had very little to say.
Now you can ask ChatGPT to produce a thought leadership piece on almost any subject and receive something that looks reasonably polished in seconds.
It will probably have a compelling headline, a clear structure, some confident assertions and a neat conclusion.
It might even contain an em-dash or two.
But if you haven't given it an original idea, some meaningful evidence or a genuinely interesting perspective, there's a good chance the result will be generic.
It will sound like an intelligent person saying something important, without necessarily saying anything important at all.
AI slop sounds empty because it is empty.
Not because AI is incapable of producing good writing, but because people are using it to make themselves sound clever without doing the work of thinking of anything clever to say.
And it's not working.
The irony is that AI is an extraordinarily useful tool for developing ideas. You can use it to challenge your assumptions, explore counterarguments, uncover gaps in your reasoning and test your conclusions.
But that requires you to engage with it intellectually.
There's a world of difference between asking AI to help you think and asking it to do the thinking so that you don't have to.
And no, you probably can't always tell when something was written by AI.
There's another claim that deserves scrutiny.
"I can always tell when something has been written by AI."
Can you, though?
You might be able to recognise the default style of an unedited ChatGPT response. Certain phrases, structures and patterns have become familiar.
But recognising those patterns isn't the same as reliably identifying AI-generated text.
AI can imitate different writing styles. It can be prompted to avoid familiar constructions. It can incorporate original research, personal experience and detailed human input. And, of course, humans can write in exactly the style people now associate with AI.
There's a revealing development here.
In August 2026, Anthropic explained how it was introducing invisible watermarks into Claude-generated text. These allow specialised systems to assess whether Claude was involved in producing a piece of writing.
The main driver is compliance with the EU AI Act. But the technical challenge is instructive.
Anthropic explicitly says the watermarks are not distinguishable to human readers.
If the origin of text were reliably obvious from its writing style, there would be far less need for technical methods of identifying it.
Of course, watermarking isn't infallible either. And none of this means that people cannot sometimes identify AI-generated content correctly.
It means we should be sceptical of anyone claiming to possess an infallible AI detector between their ears.
More often, what they're detecting is generic, formulaic writing.
And then they're assuming that AI must be responsible.
Sometimes they're right.
Sometimes they're not.
The more useful question isn't who wrote it.
It's whether it's worth reading.
When I read an article, I want to know whether it tells me something I didn't already know.
Does it challenge an assumption? Present new evidence? Offer a useful explanation? Help me make a better decision?
Those are the things that give writing value.
Whether the author used ChatGPT, Microsoft Word, a typewriter or a pencil is a secondary consideration.
A genuinely insightful article doesn't become worthless because AI helped produce it.
And a vacuous article doesn't become valuable just because a human painstakingly typed every word.
We should absolutely challenge the growing volume of generic content being pumped into the world under the banner of thought leadership.
But we should be clear about what we're challenging.
The problem isn't that people are using AI to write.
It's that they're using AI to avoid thinking.
And no amount of clever prompting, elegant prose or carefully rationed em-dashes can compensate for having nothing interesting to say.
Apparently, that's an unpopular opinion.
Or perhaps it's just another way of saying something everyone already knows.
