AI Cleans Up the "Digital Waste" It Created—Tech Companies Begin the "AI Slop" Cleanup

With the advent of generative AI, creating content on the internet has become easier than ever before.

A few seconds for text. A few dozen seconds for images. Even music and videos can be generated in a short time by individuals without specialized production environments.

This change was a significant technological innovation that expanded individual creativity. However, the "dramatic reduction in the cost of creation" has started to generate new problems.

This is the content referred to as "AI slop."

AI slop does not simply mean "something made by AI." Generally, it refers to images, texts, music, and videos mass-produced using generative AI that lack originality or informational value.

Similar articles that flood search results. Strange images that continuously flow on social media. A massive number of songs that no one knows who will listen to. Posts that are well-structured but provide little new information when read.

As the "ability to create content" exploded due to generative AI, humans have now begun to bear the "cost of finding valuable content."

And in 2026, ironically, the major tech companies that have been actively adopting generative AI began to seriously address this issue.

Spotify Deletes 75 Million "Spam Songs"

One of the figures symbolizing the AI slop issue is the "75 million" revealed by Spotify.

Spotify explained that due to the proliferation of generative AI, spam activities such as mass posting and duplicate content became easier, leading to the deletion of over 75 million spam-like songs in 12 months.

The important point is that Spotify is not trying to eliminate "music created with AI" itself.

The issue is the act of using generative technology to create a large number of songs to exploit search and recommendation systems and royalty mechanisms.

In the past, completing a single piece of music required many processes such as composition, performance, recording, and editing.

However, with the drastic reduction in production costs due to generative AI, the act of "creating a work" itself becomes almost infinitely replicable.

In a world where one artist releases 10 songs a year and a world where an automated system generates thousands of songs a day, the very design philosophy of platforms must change.

Traditional music services were built on the premise that "works are scarce."

Generative AI has broken that premise.

The structural change is the background to Spotify's advancement in spam filters, anti-impersonation measures, and increased transparency regarding AI usage.

The Era of "Reporting AI Slop" on LinkedIn

The change is not limited to music.

Business SNS LinkedIn is also strengthening its measures against AI slop.

LinkedIn has introduced a system where users can report content suspected to be low-quality AI-generated posts.

What is symbolic about this feature is the idea of "not leaving the judgment of AI content solely to the platform."

Users themselves can provide feedback on content they judge as,

"lacking substance,"

"appearing mass-generated,"

"barely reflecting personal experience or opinion."

In August 2026, it was reported that over one million people used this reporting feature within just about two weeks of its introduction.

LinkedIn is also enhancing measures against automated comments and mass postings while improving classification systems to identify low-quality content.

There is a fundamental contradiction that SNS faces here.

For a long time, SNS algorithms have grown by considering "number of posts," "engagement," and "time spent" as important metrics.

However, generative AI can optimize that game in an extreme way.

While humans could only post once a day, AI can generate 10, 100, or 1,000 posts.

Even comments can be automated.

As a result, there is a danger that "SNS for human interaction" could turn into "a place where machines generate content for machines."

LinkedIn's shift towards valuing "perspectives from real humans" indicates that the very evaluation criteria of SNS are beginning to change.

AI Captures AI

There is another intriguing phenomenon.

To crack down on AI slop, platforms themselves have begun to use AI.

In 2026, Google researchers published research on a system to detect organized synthetic content and media misuse using generative AI.

The key is not to judge each video individually but to analyze the behavior of groups of accounts that are mass-generating content.

Is the posting speed unnaturally fast?

Are multiple accounts using similar templates?

Are similar story structures or visuals being repeated?

Are there technical or behavioral characteristics common among the accounts?

By combining such information, the mass production network itself is discovered.

The research reportedly led to the handling of about 50,000 clusters, totaling about 130,000 channels over six months.

In other words, what is happening now is, in a sense, a "competition of AI against AI."

The side that creates massive content using generative AI and the side that detects mass generation networks using AI.

As generative technology advances, detection technology also advances.

The battle that spam emails and spam filters have waged for years is now about to spread across the entire internet in the form of text, images, music, and videos.

Snapchat Prioritizes "Human-Created" Content

Another noteworthy move is by Snapchat.

In July 2026, Snap announced its policy to exclude fully AI-generated videos from recommendations on its video posting service Spotlight.

On the other hand, works that use AI for editing or production assistance are not necessarily excluded.

Here, too, an important boundary is visible.

The issue is not

"whether AI was used"

but rather

"whether there is the creator's intention or originality."

This is the perspective.

Generative AI has the potential to become a creative tool like cameras, Photoshop, or synthesizers.

However, can a system that mass-generates with a single click and posts without much verification be called the same "creation"?

Platforms are now being forced to draw that line.

Complaints of "Exhaustion from Searching" on SNS

So how do users feel?

 

Looking at SNS like Reddit, complaints about AI slop are quite direct.

Particularly noticeable is the voice that says,

"I have to go through a lot of AI content before reaching the information I want."

In communities around Pinterest, complaints are repeatedly posted about how searching for illustrations, interiors, or photo references results in a flood of AI-generated images.

Some users say that even using settings to reduce AI content does not completely eliminate it.

In another Reddit community, there is a movement to ban AI slop posts and strengthen user verification due to the problem awareness that "we come to get advice from real humans, but end up reading AI-generated texts."

What becomes apparent from these reactions is not so much a rejection of AI itself, but rather frustration over "having their time taken away."

People do not want to read 100 search results.

They want to find one necessary answer.

They do not want to see 1,000 images.

They want to find 10 useful ones.

They do not want to choose from a million songs.

They want to encounter a few that suit them.

With the explosive increase in supply due to generative AI, what has become scarce is not content, but "human attention."

The Counterargument: "Don't Call It Slop Just Because AI Was Used"

However, there is also a problem if the backlash against AI slop becomes too strong.

Already on SNS,

there is a counterargument that "it's wrong to call it slop just because AI was used, without evaluating the quality of the work."

For example, in software development, even if AI writes part of the code, if humans design, test, modify, and verify safety, it should not be dismissed as low quality.

The same applies to image creation.

Using AI to consider composition and automatically generating thousands of images for posting are entirely different cases.

In music, if artists use AI as an instrument, and spam operators mass-post auto-generated songs, putting them in the same category could risk losing the creativity that should be protected.

Google's research also emphasizes focusing more on organizational and mass-produced behavior patterns than on individual creators trying new tools.

What will become important is not a simple binary choice of "AI or human."

How much human judgment is involved?

Is there unique information?

Is it valuable to someone?

Is there an attempt to exploit the system through mass production?

These will likely be the criteria.

The Essence of the AI Slop Problem is "Trust"

More serious than the increase in low-quality content is the decline in trust in the entire internet.

When you see a photo,

you wonder, "Is this real?"

When you read a review,

you suspect, "Was this written by AI?"

When you read a LinkedIn post,

you wonder, "Is this the person's experience or a generated text?"

When you watch a video,

you doubt, "Does this person exist?"

If you have to verify each one, the psychological cost of using the internet increases significantly.

Spam, copy articles, and clickbait existed even before generative AI.

But what generative AI has significantly changed is the "scale."

Creating 100 malicious sites and creating 100,000 using AI have entirely different impacts on the information environment.

If the cost of producing low-quality information approaches zero, platforms must create mechanisms to "prove quality" in return.

The Fact That "Humans Created It" Becomes a Brand

From here on, an interesting reversal phenomenon may occur.

Before generative AI, the fact that something was digital content was new in itself.

However, if AI-generated products become commonplace, conversely,

"photos taken by humans,"

"texts written by the person,"

"reviews by people who actually visited,"

"music performed by musicians,"

will become valuable.

Already on SNS, there is a movement to prominently feature "genuine," "original," and "human-created" works.

In the future, like ingredient labeling on food,

human-made

AI-assisted

AI-generated

could become common labels for content production processes.

What users seek is not necessarily "zero AI."

What they want is transparency to judge what they are viewing.

From the Era of Content to the Era of "Selection"

When generative AI first appeared, many companies emphasized the advantage of "being able to create more content than ever before."

You can create 10 times more blogs.

You can create 100 times more ad images.

You can post a large number of videos.

You can update SNS daily.

But by 2026, that very idea is beginning to reach its limit.

If everyone creates 10 times more, the total content volume in the world also becomes 10 times more.

Then, it does not necessarily mean that "those who create a lot" will win.

Rather, to avoid being buried under a massive amount of information,

what not to create.

what not to publish.

where to add human judgment