What Is AI Slop? Why There's So Much of It Online

The short version

What’s happening: AI slop is low-quality content made in bulk with AI tools. What makes something slop is the lack of effort and checking, rather than the use of AI itself.
Why now: Merriam-Webster made 'slop' its 2025 Word of the Year, and in 2026 Deezer reported that AI tracks passed half of its new uploads on peak days.
Should you care: Slop takes up room in search results and feeds, and some of it is simply wrong. A few app settings and habits can reduce how much of it you see.

What is AI slop? It’s low-quality content churned out in bulk with AI tools: filler articles, strange images, throwaway videos, soundalike songs. Merriam-Webster defines it as “digital content of low quality that is produced usually in quantity by means of artificial intelligence.” In our view, effort decides whether something counts, not the tool.

There’s so much of it because making it now costs almost nothing, while attention can still earn money. The surprise is in the numbers: slop swamps what gets uploaded far more than what gets watched. Some apps also have settings that turn it down.

What is AI slop, and what doesn’t count?

Not everything made with AI is slop, though there’s no agreed line. Kapwing’s report points out that some people use the term only for unreviewed work, some include lightly reviewed work, and some apply it to anything AI-generated.

YouTube’s rules lean towards results over tools. Its monetisation page accepts automated tools, provided the final video shows the creator’s vision and offers educational or entertainment value.

Here’s a quick self-check. It’s our own working test, not an official standard.

Ask this	Leans towards slop	Leans towards fine
Did anyone check it?	Errors, invented details, nobody answerable	A named person stands behind it
Does it add anything?	Rewords what already exists	Brings facts, skill or a view of its own
Why was it made?	To collect clicks or plays	To inform, entertain or help
How was it made?	Many near-identical versions	Each piece made on purpose

That test applies to this site as well: our standards page says AI tools may assist with research, drafting and proofreading, and that a person reads each story in full and checks each fact against its source.

How it spread

15 Jul 2025YouTube renames its 'repetitious content' monetisation policy to 'inauthentic content'.
18 Nov 2025TikTok announces a control for how much AI-generated content appears in the For You feed.
28 Nov 2025Kapwing publishes its report, including a test of the first 500 Shorts shown to one new account.
15 Dec 2025Merriam-Webster names 'slop' its 2025 Word of the Year.
21 Jan 2026YouTube's chief executive addresses low-quality AI content in his annual letter.
21 Jul 2026Deezer reports that fully AI-generated tracks passed 50% of new uploads at their June peak, while staying at 1 to 3% of streams.

Where did the word “slop” come from?

The word is older than the internet by centuries. Engadget’s report on Merriam-Webster’s choice traces it from soft mud in the 1700s to food waste, often fed to pigs, in the 1800s. From there it widened to mean rubbish.

The dictionary named it Word of the Year on 15 December 2025. Its president told the Associated Press it was “almost a defiant word” when it comes to AI.

Why is there so much AI slop?

Two forces push the same way. The first is cost: another article, image or clip now takes minutes and very little money, so a maker can afford for most attempts to fail.

The second is reward. Ad shares, royalties and affiliate links all pay for attention, whatever its quality. Put those together and volume becomes a sensible strategy for anyone who cares only about the payout.

Platforms treat this as a known problem. In a 15 July 2025 update, YouTube renamed its “repetitious content” policy to “inauthentic content”, and said this covers repetitive or mass-produced work that was already barred from earning. In his January 2026 letter, YouTube’s chief executive said the company was extending its spam and clickbait systems to reduce low-quality AI content.

Music shows a harsher version of the same incentive. Deezer says up to 85% of streams on fully AI-generated tracks in 2025 were fraudulent, meaning fake plays rather than people listening. Deezer doesn’t name those responsible, and neither can we.

How much of the internet is AI slop?

Nobody knows, because nobody has a slop detector. Most of the figures below come from tools that estimate whether a machine wrote something, and those tools say nothing about whether it’s any good. They can spot a machine’s habits, but they can’t spot a lack of care.

So read each number as what it is:

Who measured	What they looked at	Result	The limit
Deezer	New tracks delivered to its service	Over 50% fully AI on peak days in June 2026; about 90,000 tracks a day in its notes, though its opening calls 90,000 a monthly average	One service, using its own detector
Graphite	55,400 English articles from a public web archive	49.9% primarily AI in Q1 2026	No traffic data; human-edited AI drafts not assessed
Pangram	1,002,627 social posts scanned by its browser extension users	25.72% of posts over 250 words flagged as fully AI	Opt-in users, and a detector firm reporting on its own tool
Kapwing	The first 500 YouTube Shorts on one new account	104 (21%) AI-generated; 165 (33%) what it calls “brainrot”	One account, one test, by a video-tool company

These don’t add up to one answer, and they shouldn’t be averaged. Kapwing’s figure describes one feed on one occasion, so it isn’t a claim about YouTube in general. AI-written also doesn’t automatically mean slop: a useful article drafted with AI would still count in Graphite’s total, if it read as primarily AI.

Does anyone actually watch or read it?

Far fewer than the upload counts suggest. Slop is winning the upload count much more clearly than it’s winning your attention.

On Deezer, fully AI-generated music accounts for just 1 to 3% of streams, even with more than half of new uploads at peak. There’s a catch, though. Deezer keeps these tracks out of its algorithmic recommendations and editorial playlists, so that low share partly reflects its own policy. Other services with other rules could look different.

Articles show a similar gap. In Graphite’s search study, published in October 2025 using a different detector and Google results from June 2025:

86% of articles in Google’s results were human-written and 14% AI-generated
7% of number-one results were AI-generated
82% of articles cited by ChatGPT and Perplexity were human-written

Search results also draw on years of older pages, so some of that gap may simply reflect age. That’s our caution rather than a finding of the study. Graphite’s 2026 paper says it didn’t measure traffic, though its authors suspect AI articles get less of it.

Why does my feed still feel full of it?

What follows is reasoning, not a proven finding about any one platform. Feeds tend to favour what earns a reaction, and something bizarre earns one: a pause, a share, a mocking comment. To a ranking system, that disbelief may look much like interest.

A feed also shows you one item at a time, so a small number of odd posts can feel like the whole picture. You remember the strange fake and forget the ordinary clips around it. For another format built around the feed, see our explainer on microdramas.

Kapwing’s test doesn’t settle this. It found a lot of AI video on a new account, but says it can’t tell whether that came from the recommendation system or the sheer number of uploads, and that only Google could answer.

How do I see less AI slop?

The menu names below come from the sources linked, and apps change them often. We haven’t tested these steps on our own accounts, so yours may differ.

Pinterest

On a Pin in your feed, tap the ellipsis at the bottom right and pick the “see fewer” AI Pins option. It’s offered only in certain popular categories, and it works as a signal to your recommendations.
For wider control, AP described this path in January 2026: Settings > Refine your recommendations > GenAI interests, then use the toggles to see less in each category.

TikTok

Go to Settings > Content Preferences > Manage Topics and move the AI-generated content slider down.
AP gave a second route: on a For You post, tap Share > Why this video > Adjust your For You > Manage topics.
In January 2026, AP still described the slider as a test, and it wasn’t available in desktop browsers. We couldn’t confirm how widely it’s available as of November 2026. TikTok says it adjusts your feed and doesn’t remove AI content entirely.

Any feed

Use “not interested” or the equivalent when it’s offered.
Don’t comment on slop to mock it, and don’t share it as a joke. On the reasoning above, that reaction may bring you more.
Follow people and publications you trust, and read from your following list where the app has one.

The bigger risk is slop that looks right and is wrong. Before you repeat or rely on a surprising claim, check who is answerable for it and find the original source.

Is AI slop going away, or getting worse?

Neither answer fits everywhere, as of November 2026. Articles look flat: Graphite’s share has hovered near 50% since early 2025, and its authors suspect that’s because such articles do badly in search.

Music is still climbing. AP reported in January 2026 that Deezer was receiving about 60,000 fully AI tracks a day, over 39% of uploads. The June peak was far higher, although a peak-day figure and a general daily figure aren’t strictly like for like.

Platforms are mostly adding dials and limits instead of trying to clear it all out. Deezer, for example, tags AI tracks for listeners and says it will remove those used for streaming fraud and those left unstreamed for six months.

What isn’t known is how long the measurements will hold. Graphite warns that its detectors were tested on only three AI models, and that newer ones may be harder to catch. So tomorrow’s numbers may undercount.

The flood isn’t going anywhere. Your feed, at least, takes instructions.

Our take

The argument about AI slop too often stops at the tool. The better question is whether anyone cared enough to check the thing before asking for your attention. The early numbers suggest that making slop is easy and getting people to choose it is much harder, and that should reassure anyone who still makes things with care.

Common questions

What is AI slop on YouTube?

It usually means mass-produced AI videos made to collect views, such as templated clips with little story or thought behind them. YouTube's monetisation rules say AI-generated content built from generic templates, giving an impression of mass production, can't earn money.

Is AI slop allowed on YouTube?

YouTube's monetisation page doesn't treat AI use as a problem in itself. As of November 2026, it says automated tools are fine if the result shows creative vision and has value, while repetitive or mass-produced content isn't eligible to earn money. That page covers earnings, so other YouTube rules may also apply.

What is AI slop on LinkedIn?

We found no formal definition. People generally use it for polished posts with no real experience behind them. Pangram, which sells AI detectors, reported in July 2026 that its tool flagged more than 40% of long LinkedIn posts in its opt-in sample as fully AI-generated. That's a detector estimate, not proof of authorship.

What is AI slop music?

It's the same idea applied to songs: tracks generated in bulk with little human effort. Deezer says fully AI-generated tracks made up more than half of new uploads on peak days in June 2026, but only 1 to 3% of streams on its service.

Is AI slop going away?

There's no sign of that as of November 2026. Graphite's share of AI-written articles has held near 50% since early 2025, and Deezer's upload figures have kept rising. Platforms are mostly adding controls and limits, which reduce what you see without removing it.

Sources

Spotted a mistake? Tell us, or read our corrections policy.

Weekly briefing

The stories people are talking about and why, in one short email a week.

What is AI slop, and why is there so much of it?