There is a new kind of pollution choking the internet, and it does not come from factories.
It comes from prompts.
AI slop is the endless stream of low-effort, AI-generated articles, videos, images, reports, presentations and social media posts produced because machines have made creating content almost effortless. It is cheap, fast and abundant. It is also increasingly difficult to avoid.
The problem is not artificial intelligence itself. The problem is what happens when humans stop thinking and start outsourcing the thinking.
That distinction matters.
A carefully researched report assisted by AI is not slop. A student using AI to understand a difficult concept is not slop. A designer using generative tools to explore ideas is not slop.
Slop is what happens when speed replaces judgment and volume replaces value. Researchers increasingly describe it as material created with little human care for accuracy, usefulness or accountability.
And the internet is filling with it.
Scroll through YouTube, TikTok, Facebook or Instagram and you will find synthetic videos designed to harvest attention rather than communicate anything meaningful. Search for an obscure topic and you may encounter websites producing hundreds of articles that are technically readable but intellectually empty.
It is the digital equivalent of fast food.
The frightening part is that we are beginning to consume it as though it were knowledge.
The outsourcing of thought
Education may be where the consequences become most serious.
A student who asks AI to explain algebra has gained a tutor. A student who asks it to write the assignment, generate the arguments and provide the conclusion has gained something else: a shortcut around learning.
The OECD’s 2026 Digital Education Outlook warns that GenAI can improve performance on tasks without producing genuine learning when students simply outsource cognitive work. It specifically identifies risks of disengagement and “metacognitive laziness”.
That should terrify educators.
Because education is not primarily about producing a polished answer.
It is about developing the brain capable of producing one.
There is a reason some universities are now restricting AI in core learning. The University of Chicago Law School, for instance, is tightening technology use in foundational courses because educators fear that shortcuts can interfere with the intellectual struggle through which judgment is developed.
We may be creating a generation that can produce excellent-looking work without necessarily understanding it.
And then there is the workplace
AI was supposed to eliminate busywork.
Instead, it may be creating a new category of it.
“Workslop” describes AI-generated workplace material that looks polished but lacks substance: bloated reports, meaningless presentations, generic emails and code that somebody else must inspect and repair.
Research from BetterUp Labs and Stanford’s Social Media Lab found that 40 per cent of surveyed US desk workers had received workslop in the previous month, with each incident taking an average of two hours to resolve.
That is the dirty secret of AI productivity.
One person’s efficiency can become another person’s workload.
The employee who generates a 30-page AI report in three minutes has not necessarily saved the company time. If their manager spends two hours figuring out what the report actually says, the organisation has simply moved the labour around.
The International Labour Organisation’s 2026 review similarly finds that AI productivity gains are real but uneven, while warning about inequality, job quality and declining opportunities for younger workers.
The intelligence problem
There is something deeper happening.
Human intelligence develops through friction.
We read difficult things. We write badly before we write well. We make mistakes. We argue. We research. We struggle to find the right word. We solve problems without knowing the answer.
That struggle is not inefficiency.
It is training.
If AI removes every difficult cognitive task, we risk confusing the ability to obtain an answer with the ability to understand one.
That is the paradox of generative AI: it can make us extraordinarily productive while simultaneously making us intellectually lazy.
But there is another side.
AI can democratise expertise. A student in Nairobi can access a tutor at any hour. A small business can produce professional material without hiring an entire agency. A researcher can interrogate mountains of information. A programmer can prototype ideas faster.
Used properly, AI is an intellectual amplifier.
Used lazily, it is an intellectual wheelchair.
The future therefore should not be anti-AI. That would be absurd.
It should be anti-slop.
Schools should teach students how to use AI without allowing it to replace foundational skills. Companies should reward useful thinking rather than the volume of documents produced. Media organisations should disclose and aggressively edit synthetic material. Platforms should stop treating infinite content as automatically valuable.
And individuals need to recover an increasingly unfashionable habit:
thinking for themselves.
Because the biggest danger of AI is not that machines will become intelligent.
It is that humans will become comfortable being less intelligent because the machines are doing the work.
And if that happens, the most successful AI system in history may not be the one that makes machines smarter.
It may be the one that convinces humans they no longer need to think.
