Content Farms, SEO Slop, and the Long Hangover of Cheap Writing

In 2011, Google ran an algorithm update they named Panda. It targeted a specific kind of website: one that published enormous volumes of low-quality content, optimized not to be useful but to appear high in search results. These sites were called content farms, and some of them were producing tens of thousands of articles per day — written by underpaid contractors, quickly, to hit keywords.

Panda hit them hard. Within months, some of the biggest operators had lost 90% of their search traffic.

The interesting thing is that this worked, somewhat, for about a decade. The internet got measurably better. Then AI arrived, and the same dynamic came back — but at a scale that no algorithmic update has yet managed to address.

How content farms worked

The model was simple: identify a search query with decent volume and low competition, commission a 500-word article on that topic, publish it, wait for traffic. Multiply by 50,000 articles.

The writing was bad. The information was often thin or wrong. But "does this answer the question someone might type?" and "is this accurate and useful?" are different questions, and the SEO game was optimized for the first one.

Writers got paid $3-15 per article. The economics required volume. Volume required speed. Speed required compromising quality below any reasonable floor.

What AI changed

The human labor cost dropped to near zero. Everything else about the model stayed the same.

The same incentives — capture search traffic by matching keywords — are still there. The difference is that you can now produce those 50,000 articles for the cost of an API subscription rather than a small army of contractors. The quality is roughly the same. The scale is dramatically higher.

This is why search results feel so degraded right now. It's not that the algorithm is broken. It's that the incentives driving content production have been the same for twenty years, and the cost to exploit them just collapsed.

What doesn't get indexed away

The content that survives all of this — historically and probably going forward — shares certain characteristics. It's written by someone who knows more about the topic than you can learn in an afternoon of research. It takes a position that isn't just restating consensus. It contains information that couldn't have been assembled by reading other articles on the same topic.

In other words, it's reporting, or expertise, or perspective. Things that require a specific person with specific knowledge to produce. None of those things have been automated.

The writing layer is becoming a commodity. The knowing layer still isn't.