YouTube in war against AI: It’s now up to you to rate ‘garbage’ videos
YouTube appears to be seeking a new strategy to combat the growing wave of AI-generated content cluttering its recommendations. Recent reports indicate that the platform is testing a new pop-up in its mobile application that asks users if a video “resembles AI slop” or “low-quality AI,” with a response scale ranging from “not at all” to “extremely.”
While YouTube has not officially confirmed this test, several screenshots shared online suggest it is indeed taking place.
Audience-Assisted Moderation
If this test is confirmed, YouTube would be adding an extra layer to its existing systems for automatic detection and human review. The idea is straightforward: directly asking viewers to assess the impression a video leaves, not just in terms of general quality but specifically regarding “low-quality AI.”
YouTube just took a strong stance against AI slop pic.twitter.com/37pj7utWQu
— DramaAlert (@DramaAlert) March 17, 2026
This suggests that the platform’s traditional tools are still struggling to keep pace with the massive, repetitive, and often interchangeable production of these contents. This interpretation is based on the very nature of the observed test and on YouTube’s publicly stated objectives for 2026.
YouTube Acknowledges the Issue of AI Slop
This movement is not occurring by chance. In his letter outlining priorities for 2026, YouTube’s CEO, Neal Mohan, explicitly promised to “reduce the spread of low-quality AI content,” relying on systems already in place against spam, clickbait, and repetitive content.
In other words, “AI slop” is no longer just a term coined by critics of the platform; YouTube is now using it in its official communications.
Contrary to Common Belief, YouTube Already Requires Certain AI Disclosures
However, it’s important to correct a frequently misrepresented point. YouTube does not require a label for all content created with AI, but it does mandate that creators disclose realistic or synthetic content altered in a way that could be mistaken for reality. This requirement has been in place since 2024, particularly for videos depicting a person, place, scene, or event that was generated or modified by AI.
On the other hand, YouTube does not require a mention for lighter or clearly fictional uses, such as certain animations, special effects, or production aids.
The Problem Remains Vast Despite Promises
The most frequently cited figures come from a Kapwing report, as reported by The Guardian. The study claims that about 21% of the top 500 recommended videos to a new YouTube account can be classified as “AI slop,” and approximately 33% fall into a broader category of “brainrot” content, which is repetitive, low on substance, and primarily designed to grab attention.
The report also estimates that hundreds of channels built entirely on this type of content have already amassed tens of billions of views. It’s worth noting that this study comes from a private company and not from an independent academic audit, but it still provides a rough scale of the phenomenon’s magnitude.
A Broader Admission About the State of Video on the Web
The issue becomes even more concerning when it affects young viewers. This is perhaps the most opaque aspect at the moment. YouTube has not publicly explained how these user assessments would be weighted or whether they would be used for recommendation, demonetization, moderation, or training internal detection tools. In the absence of an official statement, any conclusions about the actual use of this data remain speculative.
Concerns have been raised online regarding potential reuse to train models, but these have not been confirmed at this stage.
Ultimately, this test reflects something larger than just YouTube’s strategy. For years, platforms have encouraged large-scale production, speed, optimization, and minimal friction. Generative AI has only pushed this logic to its extreme. By now asking viewers to identify what feels hollow, artificial, or worthless, YouTube implicitly acknowledges that the problem is no longer merely technical. It becomes cultural, editorial, and nearly sensory: users sense that a piece of content is “slop” even before an algorithm can accurately define it.



