The Slop Epidemic
Session 0.1 · ~5 min read
In April 2025, Ahrefs analyzed 900,000 newly published web pages. 74.2% contained AI-generated content. By November 2024, Graphite's data showed that more than half of all new web articles were primarily AI-generated. Europol projects that 90% of online content will be synthetically generated by 2026.
These are not fringe estimates. They come from SEO firms, law enforcement agencies, and researchers who track web content at scale. The internet is being flooded, and most of what is flooding it is garbage.
The Numbers
The growth trajectory is steep. In late 2022, AI-generated articles made up roughly 10% of new web content. By mid-2024, that figure crossed 40%. By early 2025, it was the majority. The tools got cheaper, the barriers vanished, and the volume exploded.
| Year | Estimated AI Content (New Pages) | Key Driver |
|---|---|---|
| 2022 | ~10% | ChatGPT launch (Nov 2022) |
| 2023 | ~25-30% | GPT-4, cheap API access, SEO farms |
| 2024 | ~50% | Subscription AI tools, one-click generators |
| 2025 | ~74% | Cost collapse, bulk publishing workflows |
| 2026 (projected) | ~90% | Full automation of content pipelines |
Note the distinction Ahrefs made: while 74% of new pages contained AI content, only 2.5% were "pure AI" with zero human editing. The rest fall on a spectrum from light AI assistance to heavy AI generation with cosmetic human touch-ups. That spectrum matters, because "contains AI content" and "is AI slop" are not the same thing.
What Slop Actually Looks Like
The term "slop" entered mainstream usage in 2024 to describe AI-generated content that nobody asked for, nobody needs, and nobody benefits from reading. It is the text equivalent of spam calls. It exists because producing it costs almost nothing and because, in aggregate, it generates just enough ad revenue to justify its existence.
Slop has recognizable characteristics. It hedges constantly ("It's important to note that..."). It uses superlatives without evidence ("This powerful approach..."). It structures everything as lists because lists are easy to generate and require no argumentative throughline. It opens with throat-clearing ("In today's fast-paced world...") and closes with platitudes ("The future is bright for those who..."). It sounds like it was written by a committee that has never had a specific opinion about anything.
Slop is not AI content that happens to be bad. Slop is content produced with no quality standard, no editorial judgment, and no accountability. The AI is the tool. The absence of standards is the problem.
How We Got Here
The path from "AI can write text" to "the internet is drowning in garbage" took about 18 months. The mechanism is straightforward.
($0.50-3 per 1M tokens)"] --> B["Bulk publishing tools emerge"] B --> C["Content farms scale to
500+ articles/month"] C --> D["Volume floods every niche"] D --> E["Average content quality
collapses"] E --> F["Readers lose trust
in written content"] F --> G["Genuine expertise gets
buried under noise"] G --> H["The slop epidemic"]
Each step follows logically from the previous one. When generating an article costs less than a dollar, the rational economic move is to generate as many as possible. When your competitor publishes 500 articles a month, you publish 600. Nobody stops to ask whether any of those articles are worth reading, because "worth reading" is not the metric. Volume is the metric. Rankings are the metric. Ad impressions are the metric.
The Quality Collapse
Originality.ai has been tracking AI content in Google search results since 2023. Their ongoing study found that AI content in search results peaked at about 19.5% in mid-2025 before Google's ranking adjustments brought it down slightly. But that 19.5% represents what ranks, not what exists. The total volume of AI content published is far higher. Most of it never ranks at all. It sits in the long tail of the web, polluting niche queries and eroding the baseline quality of information.
The collapse is not uniform. High-stakes content (medical, legal, financial) has been somewhat protected by Google's stricter quality requirements for "Your Money or Your Life" (YMYL) topics. But informational content, product reviews, how-to guides, and general knowledge articles have been hit hard. These are exactly the categories where AI generation is cheapest and where editorial oversight is thinnest.
Why This Matters to You
If you are reading this course, you probably care about producing content that is worth something. Maybe you write for a living. Maybe you run a business that depends on content. Maybe you are building a publishing operation and want to use AI without contributing to the problem.
The slop epidemic creates both a problem and an opportunity. The problem: your work competes for attention in an environment where noise has overwhelmed signal. The opportunity: the bar is on the floor. Anyone who produces content with genuine standards, real expertise, and actual editorial judgment stands out simply by not being garbage.
This course is about building the systems, workflows, and quality controls that let you use AI as production infrastructure while maintaining standards that most operations have abandoned. The goal is not to avoid AI. The goal is to use it without losing your mind, your voice, or your readers' trust.
Further Reading
- 74% of New Webpages Include AI Content (Study of 900k Pages) (Ahrefs, 2025)
- Amount of AI Content in Google Search Results (Originality.ai, ongoing study)
- Over 50 Percent of the Internet Is Now AI Slop (Futurism)
- Experts: 90% of Online Content Will Be AI-Generated by 2026 (The Living Library)
Assignment
- Find 5 pieces of AI-generated content in the wild: articles, product descriptions, social media posts, or email newsletters.
- For each one, write 2-3 sentences explaining exactly what makes it feel like slop. "It sounds fake" is not specific enough. Identify the specific patterns: the hedging, the generic advice, the absence of lived experience, the formatting tics.
- Organize your findings in a table with columns: Source, Content Type, Specific Slop Markers, and What a Human Version Would Include.