How to Block AI Videos on YouTube (5 Tools That Actually Work)

Install a browser extension or blocklist built specifically to detect and hide AI videos. YouTube's own tools were never designed to filter recommendations, only to label them. Your options range from machine-learning classifiers to manual, self-maintained channel lists, and the right pick depends on how much control you want over your own feed.

That distinction matters more than it sounds. YouTube has spent two years building disclosure infrastructure for AI content. It has not built anything that stops that content from reaching you.

Why Doesn't YouTube's Labeling System Block AI Videos?

In March 2024, YouTube introduced a Creator Studio setting requiring creators to disclose when realistic content is "meaningfully altered or synthetic," including generative AI.¹ Most videos get a small label tucked into the expanded description. More sensitive content gets a prominent overlay.¹

In May 2026, YouTube went further, automatically detecting undisclosed photorealistic AI content and moving labels to more visible spots, including an overlay on mobile.² But the company was explicit about what this does and doesn't do. As YouTube put it, "these labels alone do not affect how our videos are recommended or whether they can earn money."² Translation: the label is a transparency feature, not a filter. A flagged AI video can still land at the top of your homepage.

The manual alternative, marking videos "Not Interested" or hitting thumbs-down, fares no better. Users on YouTube's own community help forums report doing this for months with zero change in what gets recommended. The feedback loop doesn't seem to weight AI content differently than anything else. So the burden of filtering falls entirely on the viewer.

How Much AI Slop Is Actually in YouTube's Recommendations?

The scale explains the frustration. A Kapwing study tracking the first 500 videos recommended to a brand-new YouTube account found 104 of them, about 21%, classified as AI slop.³ Fold in adjacent "brainrot" content and that figure climbs as high as 33%.³

This isn't a fringe phenomenon confined to a handful of small channels. Kapwing identified 278 prominent YouTube channels made entirely of AI slop that have collectively racked up roughly 63 billion views and 221 million subscribers.³ For comparison, a related study found TikTok's new-account recommendation rate for AI slop sits at 59%, nearly three times YouTube's share. YouTube has taken the more aggressive enforcement stance of the two platforms, terminating 16 channels with a combined 35 million subscribers and nearly 5 billion lifetime views under its inauthentic-content policy in January 2026.⁴

The economics explain why the flood keeps coming. NPR reported on one AI-generated cat video channel that pulled in $9,000 in a single month and has amassed nearly 500 million cumulative views across roughly 600,000 subscribers.⁵ When a creator can generate that kind of return without filming anything, platform policy alone won't stem the supply. Viewers have to build their own defenses.

5 Tools for Blocking AI Videos on YouTube

A small ecosystem of independent tools has grown up around exactly this problem. None are official YouTube products. All exist because the platform's native controls fell short. Ranked roughly from most automated to most manual:

1. Clarity: an ML-trained AI video hider

Clarity is a Chrome extension built by two developers who got tired of AI slop clogging their recommendations. It hides AI-generated and content-farmed videos from your feed using a detection model trained on more than 10,000 manually labeled AI channels.⁶ That's a meaningful training set for a community project, and it's what separates Clarity from a keyword filter: it's trying to recognize patterns in AI content itself, not just match known bad channels.

The tradeoff of any ML classifier is opacity. You're trusting someone else's model to draw the line between synthetic and real, and you won't always know why a given video got flagged. For viewers who want something that mostly works out of the box, that tradeoff is usually worth it.

2. AiBlock, DeSlop, and Slop Block: community blocklists

A cluster of tools take a different approach: crowdsourced blocklists maintained by users instead of trained models. Firefox's AiBlock for YouTube lets you right-click any channel to add it to a blacklist, warnlist, or whitelist, pulling from a community-powered list others contribute to.⁷ The GitHub project DeSlop works similarly, fetching a blocklist and stripping matching videos from your feed.⁸

What's notable about DeSlop is its stated reason for staying off the Chrome Web Store entirely. Its creator wrote bluntly, "I'm not doing Google's age verification and paying the fee just to get denied because they love AI slop."⁸ Agree with that read of Google's incentives or not, it's a reminder that these tools exist outside the platform's blessing, not with it. Slop Block, in the same category, follows the same crowdsourced-reporting model. The strength of this group is transparency: you can see exactly which channels are blocked and why, and edit the list yourself.

3. AI Content Shield: blocks AI Overviews too

AI Content Shield widens the aperture beyond YouTube. The Chrome and Firefox extension blocks AI-generated and AI-voiced videos on YouTube while also hiding Google's AI Overviews and AI Mode in search results. That makes it less a YouTube-specific fix and more a general attempt to strip AI-generated content out of everyday browsing, a useful reminder that the slop problem isn't contained to one platform. If your frustration extends past your video feed and into your search results, this is the tool built for that broader complaint.

4. YTBlock: manual, customizable channel blocking

YTBlock is the DIY option. Available for Chrome, Firefox, and Edge, it lets you build and maintain your own blocklist of channels, and it ships with a default list drawn from community-sourced blocklists, including one originally built to catch AI music channels.⁹ Nothing here is automated detection. You decide what goes on the list, and you can edit it whenever you like.

That manual model is slower to set up than a trained classifier, but it puts the decision entirely in your hands rather than a vendor's. It's the same instinct behind treating any software you rely on as something you actually control rather than something a platform grants you access to, an idea explored at length in The Hardware and Software Ownership Renaissance. A blocklist you own and edit yourself doesn't disappear if a developer abandons a Chrome Web Store listing.

5. Slop Evader: rewinds the web to before ChatGPT

Slop Evader takes the bluntest approach of the five. Rather than trying to identify AI content, it filters search and video results so only content published before November 30, 2022, the date ChatGPT launched publicly, shows up.¹⁰ It works across YouTube, Google, Reddit, Quora, Pinterest, and Stack Exchange, effectively letting you browse a version of the web frozen before generative AI became commercially available.¹⁰

It's a crude filter by design. You lose three-plus years of legitimate human-made content along with the slop. But for anyone who wants a guaranteed floor rather than a probabilistic guess, a hard date cutoff is the one method that can't be fooled by a better AI model next year.

The Real Fix Is Owning Your Feed, Not Renting It

What these five tools share isn't a technique. It's a posture. None wait for YouTube to fix the problem. Each assumes that if the platform won't give you real control over your recommendations, you build or adopt something that does.

That's a small, specific instance of a pattern showing up across software more broadly: people deciding that renting someone else's black box, whether that's a recommendation algorithm or a SaaS subscription, isn't good enough when the thing actually matters to them. Any business or workflow built entirely on someone else's platform access lives at the mercy of a policy change, as the Nitter and XCancel takedowns made clear. A blocklist you maintain yourself, or an extension whose logic you can inspect on GitHub, behaves less like a rented service and more like an asset. It doesn't change behind your back. It doesn't get monetized against your interests.

YouTube isn't going to build a real filter for AI slop anytime soon, because its labeling system was never designed to reduce that content's reach.² Until that changes, the tools on this list are the closest thing to owning your own feed.