AI Slop

For a few years now, there have been programs that generate texts, images, or videos at the push of a button. When someone uses these to produce large volumes of content that nobody really needs, the result is called AI Slop. "Slop" is English and roughly means slush or pig feed. The term is therefore an insult, not a neutral category. What's meant, for example, are guide articles that merely tick off search terms, made-up news sites, or images with six fingers that rack up millions of clicks on social networks. What matters is not that a machine was involved, but that quantity was put above quality.

Why this makes the internet harder to search

Search engines, recommendation lists, and timelines on social networks operate on one assumption: creating content takes effort. Anyone who writes a good text invests hours. This assumption no longer holds. A single operator can now publish thousands of articles per day. The cost of doing so is in the range of a few cents per piece.

This shifts the ratio of signal to noise. Anyone who asks a question about a medication, a trip, or a tax rule increasingly ends up on pages that sound plausible but that nobody has checked. This is especially tricky when it comes to health and money. Wrong information there is not just believed, it costs people something concrete.

There is also a longer-term problem. New AI models learn from texts and images found online. If more and more of that material is machine-generated, the next generation of models partly learns from the outputs of their predecessors. Researchers call this model collapse: quality slowly declines because errors and quirks are amplified across generations.

How the mass production works

The typical process is simple and resembles an assembly line. First, a program searches for queries that many people ask and for which advertisers pay. Then a language model writes an article for each of these queries. A language model is a program that has learned to predict the next word each time, and in this way writes entire texts. Finally, a script publishes everything automatically and adds ad placements.

For images and videos, the logic is the same. An image generator creates an emotionally charged scene, such as a child with a homemade sculpture or a made-up flood disaster. Such motifs generate a lot of reactions, and reactions bring reach. The operators earn money through the platforms' ad revenue sharing or later resell the accounts they've built up.

An important distinction needs to be made here. AI Slop is not the same as a hallucination. A hallucination is a single made-up detail in a model's text, such as a wrong year. AI Slop, by contrast, describes the business practice behind it. A text can be factually correct and still be slop, if it only exists to fill an ad slot.

Where you encounter this content every day

The effect is most visible in searches for everyday problems. Recipes, repair guides, product comparisons, and travel tips are heavily affected. Online shops are also full of it: books with made-up content, mushroom guides with dangerously wrong information, T-shirts with generated designs. On music platforms, providers upload thousands of artificial tracks to collect fractions of a cent per play.

In finance and tech news, the term comes up for a different reason. Platforms like Google, YouTube, Spotify, and Amazon have to react, because otherwise their recommendations become useless. They change ranking rules, require labeling, or delete accounts on a large scale. This is relevant for investors, because ad revenue depends on users trusting the results.

A common misconception is that slop can always be recognized by poor quality. That was true for early images, where six fingers or unreadable text were typical tell-tale signs. Today's systems make these mistakes less often. Other signals are more useful: no traceable author, an imprint without an address, hundreds of articles within a few days, many texts without concrete sources. Checking two independent sources on important topics is the most reliable way to avoid the problem.