AI Slop
Low-quality AI-generated content produced because it's cheap rather than because anyone wanted it — an economics problem that gets mistaken for a technology problem.
When not to use it
- As a synonym for AI-generated. The word is about indifference to the reader, not about the tool — commissioned, checked, accountable AI output isn't slop.
- To dismiss work you haven't read. It's become a cheap way to discredit anything, which is its own kind of laziness.
- As a technical claim about model degradation. That's model collapse, it's a different argument, and it doesn't say what people think.
Reach for something else instead
- Content farming names the same behaviour with a decade of history and no confusion about the tool.
- Model collapse is the technical claim, if that's what you mean.
- Spam is the honest precedent — same economics, same incentives, and the same reason detection was never the fix.
Sources & further reading
- Willison (2024), Slop is the new name for unwanted AI-generated content — the reference that fixed the term's meaning as a publishing decision rather than a property of output.
- Shumailov et al. (2024), AI models collapse when trained on recursively generated data — the feedback mechanism, under the replace regime.
- Gerstgrasser et al. (2024), Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data — why the collapse argument doesn't transfer to how anyone actually trains.
Primary sources, listed so you can check the claims on this page rather than take them on trust.
Where people go wrong
- Treating it as a property of AI text. It's a property of publishing without caring whether it's read — a human tradition AI made cheaper.
- Citing model collapse to argue slop will poison future models. That result needs the replace regime; real pipelines accumulate and filter.
- Measuring it with AI detectors. The tools don't work, so the widely-quoted "X% of the web is AI" figures are built on a method with a published failure mode.