What is AI slop? Why low-effort AI content is everywhere

AI slop is low-quality text, images, video or audio produced in bulk with artificial intelligence and pushed at people who didn't ask for it, usually to win clicks, views or advertising money. What makes it slop is not that a machine helped make it, but that nobody cared whether it was true, useful or good. Merriam-Webster named "slop" its word of the year for 2025.

In short

• Slop is to AI content what spam is to email: unwanted, mass-produced, and made because it costs almost nothing.
• It spreads because generating content became nearly free while platforms still pay out for attention.
• The line that matters is effort and care, not whether AI was used: AI-assisted work that someone checked and stands behind is not slop.

Where does the term "AI slop" come from?

"Slop" once meant soft mud, then food waste fed to pigs, then anything of little value. As AI image generators spread in 2022, it became slang on forums such as 4chan and Hacker News for the flood of machine-made filler. The programmer and writer Simon Willison brought it into wide use in a blog post on 8 May 2024, arguing that just as "spam" became the word for unwanted email, the field needed a word for unwanted AI content. His definition: content that is "mindlessly generated and thrust upon someone who didn't ask for it."

By the end of 2025 the word was mainstream. On 15 December 2025 Merriam-Webster named "slop" its word of the year, defining it as "digital content of low quality that is produced usually in quantity by means of artificial intelligence". Australia's Macquarie Dictionary chose the full phrase "AI slop" for the same year.

What does AI slop look like?

It takes every form that generative AI can produce.

• Images. Surreal pictures made to farm likes and shares on social media. In 2024, Facebook filled with images of "Shrimp Jesus", a figure of Jesus made of shrimp, among many stranger variations.
• Articles. Websites of thousands of machine-written pages aimed at search keywords, recipe blogs with invented recipes, and product reviews by no one who used the product.
• Books. AI-written guides sold under made-up author names, including mushroom-foraging guides that experts warned could get readers poisoned.
• Video and music. Channels that post narrated AI videos many times a day, and AI-generated tracks uploaded to streaming services in bulk.
• At work. Reports and emails that look finished but say little, leaving the reader to do the real work, sometimes called "workslop".

What they share is that volume matters more than quality. Some slop is harmless filler; some misleads, such as fake disaster photos shared as real. When it impersonates real people, it overlaps with deepfakes.

Why is AI slop everywhere? A worked example

The short answer is that the cost of making content fell close to zero while the pay for attention didn't. Here is how the arithmetic works for a website, using illustrative round numbers rather than any real site's figures.

1. The human way. A freelance writer charges $50 for a short article. 10,000 articles cost $500,000.
2. The slop way. An AI model writes a similar-looking article for about 1 cent in computing costs. 10,000 articles cost $100, and take an afternoon to publish.
3. The traffic. Most of those pages are found by almost no one. Say each gets just 200 visits over its life: 10,000 × 200 is 2 million visits.
4. The pay. Display advertising might pay around $1 per 1,000 visits. 2 million visits earn about $2,000.

The human-written site would lose $498,000. The slop site makes a profit of about $1,900 on $100, even though each page is nearly worthless. Slop doesn't need any single piece to succeed; it only needs to be cheap enough that a trickle of attention, multiplied by thousands, pays. The same logic drives social media accounts paid by views and books sold by the thousand at a dollar or two each.

Platforms' recommendation systems add to it. They reward whatever people click, watch or react to, and a strange AI image that makes people stop scrolling is rewarded like anything else.

What are platforms doing about it?

Platforms have mostly targeted the bulk and the deception, not AI as such. Google's spam policies treat "scaled content abuse", many pages made mainly to rank in search, as spam "no matter how it's created", and give as an example using generative AI to produce many pages "without adding value for users". YouTube, in a policy update dated 15 July 2025, renamed its rule on repetitive videos to "inauthentic content" and said channels that earn money should not post content that is "mass-produced, generic, repetitive, or manipulative". Amazon limited how many books one account can self-publish in a day. Whether these rules keep up with the volume is another matter.

There is also a cost for AI itself. A 2024 study in Nature found that models trained repeatedly on AI-generated output degrade, losing the rarer, more varied parts of what they learned, an effect its authors called model collapse. A web full of slop is worse material for training the next models.

How can you tell if something is AI slop?

No single sign proves it, and AI detection tools are unreliable. A combination of these is a good reason for doubt:

• No one stands behind it. No named author, no way to contact anyone, or an author with no history before this year.
• Generic all the way down. Smooth prose that never gives a specific number, place, name or first-hand detail.
• Wrong in ways a person would notice. Hands with too many fingers, garbled text in images, a recipe whose steps don't match its ingredients, facts that are confidently made up, which is how AI hallucination looks in the wild.
• Too much, too fast. A site, channel or author publishing dozens of pieces a day on unrelated topics.
• Made for the reaction. Images built to provoke awe, pity or outrage, with captions such as "Why don't pictures like this ever trend?"

Is all AI-generated content slop?

No. The difference is in how it was made and whether anyone answers for it.

	AI-assisted work	AI slop
Purpose	To inform, help or make something good	To fill space and catch attention
A person's role	Directs, checks facts, edits, takes responsibility	Presses a button, publishes, moves on
Sources and specifics	Real, checkable, often first-hand	Vague or invented
Volume	Limited by how much can be checked	Limited only by cost
Honesty about AI	Says how AI was used	Usually hides it

Still argued

Where the line falls is disputed. Some writers and artists consider any work made mainly by AI to be slop, because it takes the place of human craft and was trained on people's work without their consent. Others argue that the tool doesn't matter, only the result: a well-checked AI-assisted guide is useful, and plenty of human-made content is careless filler. A third view is that the label is applied unevenly, to whatever people dislike. All sides agree that volume without care is the core of the problem.

Slop is a side effect of how cheap artificial intelligence has made content. The useful habit is the same one that works for spam: look for who made it, what they knew, and whether anything in it can be checked.

Check yourself

• Without looking back: why does slop make money even when almost every individual piece fails?
• What do Google's and YouTube's rules target, if not the use of AI itself?
• What separates AI-assisted work from slop, in two or three words?

Sources

• Slop is the new name for unwanted AI-generated content — Simon Willison, 2024
• Merriam-Webster's 2025 word of the year is slop — Euronews, 2025
• Spam policies for Google web search — Google Search Central
• YouTube channel monetization policies — YouTube Help, 2025
• AI models collapse when trained on recursively generated data — Ilia Shumailov and others, Nature, 2024
• AI slop — Wikipedia