AI Slop
AI slop is low-quality digital content produced in large quantities with generative AI, usually with little human care and often to win attention, ad revenue, or search traffic. Merriam-Webster, which named "slop" its 2025 Word of the Year, defines it as digital content of low quality that is produced usually in quantity by means of artificial intelligence. Macquarie Dictionary, which picked "AI slop" for 2025, adds that it often contains errors and is not requested by the user. The term describes quality and intent, so not all AI-assisted content is slop.
Low-quality AI content made in bulk, usually for clicks rather than for readers.
Category: Generative AI
Where the term came from
"Slop" as a word for unwanted AI output circulated in online forums before it reached the mainstream. A June 2024 news report traced it to communities including 4chan, Hacker News, and YouTube comments. Developer Simon Willison gave it a push on May 8, 2024, arguing that slop should become the word for unwanted AI-generated content, just as spam became the word for unwanted email. He has said he did not coin it.
| Year | Recognition | Detail |
|---|---|---|
| 2024 | American Dialect Society | "AI slop" was a nominee in its Word of the Year vote |
| 2025 | Merriam-Webster | "Slop" named Word of the Year, announced in December |
| 2025 | Macquarie Dictionary | "AI slop" named Word of the Year |
| 2025 | American Dialect Society | "Slop" chosen, with the AI meaning often implied |
What counts as slop, and what doesn't
The boundary is fuzzy, but these comparisons help. The real test is whether a person took responsibility for quality, not whether AI touched the work.
| Type | Main intent | Example | Is it slop? |
|---|---|---|---|
| AI-assisted work | Help a person do better work | A writer uses AI to outline, then checks and rewrites | No |
| AI slop | Win attention cheaply and at scale | Hundreds of near-identical recipe posts with invented tips | Yes |
| Workslop | Look productive at work | A polished AI report with no real substance | A workplace form of it |
| Spam | Push unwanted messages | Bulk promotional email | Overlaps when the spam is AI-made |
| Deepfake or disinformation | Deceive about specific people or events | A fake video of a politician | Different harm; can also be low quality |
How much slop is there?
Nobody has a reliable global number. These commonly cited figures are useful snapshots, as long as you read their limits.
| Study | Finding | How to read it |
|---|---|---|
| Kapwing, November 2025 | 21% of the first 500 Shorts shown to one new YouTube account were AI slop; 33% were "brainrot" | One fresh account, not a platform-wide share |
| Graphite, 2025 | About half of new web articles in its sample of 65,000+ were AI-written | Measures AI writing, not poor quality, and detectors make errors |
| BetterUp Labs and Stanford, September 2025 | About 40% of 1,150 full-time US workers received workslop in the past month | A self-reported survey |
A useful twist from Graphite: Graphite also reported that most AI-generated articles do not appear often in Google results or ChatGPT responses. Publishing is cheap, but being seen is not. A rising share of what gets published does not mean the same share of what people read.
Why it exists: the economics
Slop follows from a simple calculation. When output costs almost nothing, even tiny earnings per item can turn a profit if volume is high.
\text{Profit} = n \times (\text{revenue per item} - \text{cost per item})
Here is an illustrative example, not real data. Suppose a page earns $0.40 on average and costs $0.05 to generate and publish. A site with 5,000 such pages makes $1,750, while a human writer at $100 per article would lose heavily on those same pages. Quality loses this contest unless platforms change the payoff, which is why demonetization matters more than labels.
How platforms are responding
- Demonetization: YouTube's inauthentic content policy targets mass-produced, repetitive videos in its Partner Program.
- Channel removals: In January 2026, reports based on Kapwing's data said more than a dozen large AI slop channels had been removed or hidden from YouTube, though counts vary.
- User controls: Platforms are adding tools that let people filter or flag AI content, after labels alone proved easy to avoid.
- Dictionary-level awareness: A shared word helps too. People can name the problem, then report and avoid it.
Enforcement has side effects: Crackdowns can catch legitimate creators. Coverage of YouTube's purge reported collateral damage for faceless human channels, because automated systems judge signals rather than intent.
How to spot slop, and how to avoid making it
There's no reliable single tell, and AI detection makes mistakes. Look for a cluster of signs: generic claims with no sources, no named author or real experience, confident errors, repeated sentence patterns, images with broken details, and a page that never quite answers your question.
| If you use AI to create | Do this |
|---|---|
| Facts | Verify every claim against a primary source |
| Point of view | Add your own experience, data, or judgment |
| Length | Cut filler and answer the question directly |
| Accountability | Put your name on it only if you would defend it |
| Volume | Publish less, and make each piece worth reading |
Honest caveats
- "Slop" can become a lazy insult. People sometimes use it for any AI-related work they dislike, including careful AI-assisted work.
- Humans made low-quality content long before AI. Content farms already existed. Generative AI lowered the cost and raised the volume.
- Measurement is weak. Studies rely on detectors, small samples, or surveys. Most public numbers come from companies with a commercial interest in the topic.
- The long-term training risk is unproven at web scale. Research on model collapse shows why recursively training on AI output is a concern, but how much slop affects current models is unclear.
Frequently asked questions
What does AI slop mean? It means cheap, low-quality content mass-produced with AI, such as bland articles, strange images, or auto-generated videos in your feed.
Is all AI-generated content slop? No. Slop is defined by low quality, mass production, and lack of care. Carefully checked AI-assisted work is not slop.
Who coined the term? No single person is credited. It began as slang in online communities, and Simon Willison helped popularize it in 2024.
What is workslop? A September 2025 Harvard Business Review article from BetterUp Labs and Stanford researchers used it for AI-generated work content that looks like good work but lacks the substance to advance a task.
Does Google penalize AI content? Not for being AI-made alone. Google says that using generative AI to create many pages without adding value for users may violate its spam policy on scaled content abuse, and it advises fact-checking AI content before publishing.
Using AI well is the opposite of producing slop. In LumiChats you can draft with a model, check it against a second one, and tie answers to your uploaded sources before you share anything.