AI slop is content made with generative AI and published without enough human judgment to be worth reading. Carol Roderick, PhD explains why the common definition, built around volume and low effort, misses the version that damages expertise-led businesses.
Nobody thinks they are the one making it.
That is the strange thing about the phrase. Every business owner who says AI slop is describing someone else. Someone else's feed, someone else's newsletter, someone else's product listing. The definition in circulation makes that easy, because it points at volume. Mass-produced. Churned out by the thousand. If you are not running a content farm, the word does not seem to be about you.
It is about you.
What is AI slop?
AI slop is content made with generative AI that gets published without enough human judgment to be worth reading. That is the whole definition. The common version adds low effort and high volume, and those do describe the most visible cases, but they are symptoms rather than the thing itself.
The volume framing spread because volume is what you can see from a distance. Fake reviews. AI-generated news sites. Feeds full of things that never happened. Most of the coverage settled on a definition built around quantity, and the dictionaries followed.
Quantity is not what makes it slop. The absence of judgment is what makes it slop.
The version that shows up in careful professional work
Here is the test that separates the two.
Take something you published recently that AI helped write. A client email, a section of a proposal, a post. Now ask what in it could only have come from you. Not what sounds like you. What could only have come from you: a position you hold that a competent competitor would not, a distinction you draw because you have watched the thing go wrong, a number from your own work.
If the answer is nothing, you published slop. It does not matter that you made one piece rather than a thousand. It does not matter that you spent forty minutes on the prompt. What went out under your name was the average of everything the model had read on that topic, tidied up.
That is the expensive version. A content farm publishing slop loses nothing, because it never had a reputation to spend. A consultant, a lawyer, an advisor publishing slop is spending the one thing clients are actually paying for.
Why this happens to people who are being careful
Effort feels like protection. It is not.
When you ask AI to write something without giving it anything specific about you, the model returns the middle of everything it has read. That middle is competent. It is readable. It is also interchangeable with a thousand other businesses in your category, because it was assembled from a thousand other businesses in your category.
Working harder on the prompt does not add your judgment to that. It gets you a cleaner version of the middle. The text improves and the problem stays exactly where it was, which is why so many people who care about their work are surprised to find themselves producing slop.
Running the result through an AI humanizer tool does not help either, for the same reason. Humanizer tools rewrite words without restoring judgment. The text comes back different and it still belongs to nobody.
How to tell before you publish
Read what you are about to send and look for the sentence a competitor could not have written. Not a better sentence. A sentence that required your specific experience to exist.
If you find one, you have something. If you read the whole thing and every line could sit comfortably under anyone's name in your field, you are about to publish slop, and the fact that it reads well is what makes it dangerous. Bad writing gets ignored. Fluent, empty writing gets read, and it teaches the reader that there is nothing here they could not get elsewhere.
What actually fixes it
The fix is upstream of the writing. AI produces the average because the average is the best available answer when it knows nothing specific about you. Give it your positions, your standards, your clients, your examples of your own writing, and the average stops being the easiest thing for it to reach for.
This is not a better prompt. It is standing context that lives inside how you run AI, so it is there every time instead of retyped at the start of each conversation. Pamela Johnston, who runs Totally Toronto Art, now has that context built as a standing project inside Claude, and the writing that comes out of it sounds like her. That case study is here.
The work is not glamorous and it is the whole difference. Slop is what a machine produces when nobody gave it anything to be specific about. Being specific is the job.
If your AI writing reads well and still feels like it belongs to nobody, the problem started before the writing did. The AI Slop Stopper is the system for fixing it at the source, and the waitlist is open. Or if you would rather look at where you are first, the AI Reality Check is free and takes five minutes.