⏱️ Lectura: 11 min
YouTube has labeled AI-generated or AI-altered videos with the “Made with AI” tag since 2024, but that notice doesn’t clean up your feed: it flashes for an instant before you keep scrolling, and the video stays right where it was. Weedout, a new macOS app, takes that same official label and uses it to remove AI videos from your feed, your search results, related videos, and Shorts before you even see them.
📑 En este artículo
The app costs $1.99, a one-time payment with no subscription, and runs on macOS 13 or later, according to its official launch page.
TL;DR
- Weedout is a native macOS 13+ app that costs a one-time $1.99, no subscription, available on the Mac App Store.
- It filters out videos labeled “Made with AI” from YouTube’s feed, search, related videos, and Shorts.
- Detection relies exclusively on YouTube’s official disclosure label, with no heuristics of its own.
- Dim mode fades AI videos out instead of removing them, so you can verify the filter before trusting it.
- It can automatically skip AI-labeled Shorts in the player, as a configurable option.
- It doesn’t create accounts, doesn’t track the user, and all processing runs locally on the Mac, with no external server.
- The developer, under the alias masteranza, receives email reports of undetected AI videos ([email protected]).
What happened: an app that removes AI videos before you see them
Weedout’s launch addresses a specific problem: YouTube already requires creators to disclose when a video was altered or generated with artificial intelligence, but the platform only shows a small notice on the thumbnail or in the description. The video still shows up in the feed, in search results, and in Shorts exactly like any other.
Weedout takes that same signal and turns it into an active filter. According to its official page, the app operates across five YouTube surfaces: the main feed (home), search, related videos, playlists, and Shorts. When it detects the “Made with AI” label in any of those places, it applies whichever mode the user has chosen: removing the video entirely or just dimming it.
Distribution is straightforward: it’s sold as a one-time $1.99 payment on the Mac App Store, with no subscription tier or freemium version. The developer identifies themselves under the alias masteranza and offers direct email support ([email protected]) for anyone who finds an AI video that slipped through undetected, explicitly asking for the video’s link to review the case.
Context and background: why YouTube labels AI content
YouTube’s “altered or synthetic” content disclosure policy didn’t originate with Weedout: it’s a platform rule that requires creators to flag their videos when they show events that never happened, real people saying or doing something they never did, or realistic scenes generated entirely with AI. YouTube has enforced it since 2024 and has tightened it over time, including automatic labeling for certain photorealistic content.
The underlying problem is the same one facing the entire AI-generated video industry: disclosure depends on the creator themselves enabling it. If a channel skips the “altered or synthetic content” checkbox when uploading a video, no external tool that relies on that label will detect it, Weedout included. The app acknowledges this plainly on its own page: “Unlabeled slop is out of scope (for now)”.
This limitation isn’t a minor detail. Earlier tools for filtering unwanted content, like ad blockers or early spam filters, had to choose between two paths: build their own classifier (with its own false positives and negatives) or rely on a signal that already exists on the platform. Weedout chose the second option, prioritizing precision over coverage.
Technical details and performance
Although the developer didn’t publish the app’s source code or a line-by-line breakdown of its architecture, the described behavior (detecting an AI-labeled element in YouTube’s DOM and hiding or dimming it on the spot) is a well-known pattern in browser extensions and userland scripts. The following table summarizes the two modes Weedout offers, according to its official documentation:
| Mode | When to use it | Advantage | Limitation |
|---|---|---|---|
| Remove (filter) | When you already trust the filter and want a clean feed | The AI video disappears completely from the grid | You can’t verify if the filter got it right without manually searching for the video |
| Dim | While you’re testing the app or want to audit each detection | You see the video “withered” and can confirm the label before trusting the filter | The video’s space still takes up room in the feed |
The conceptual flow, whatever the actual implementation behind Weedout is, can be summarized like this:
flowchart TD
A["New video in the feed"] --> B{"Has the Made with AI label"}
B -->|"Yes"| C["Weedout applies the chosen mode"]
B -->|"No"| D["The video shows normally"]
C --> E["Remove: the video disappears"]
C --> F["Dim: the video fades out in place"]To understand the mechanics without needing Weedout’s source code, you can replicate the same idea with a userscript. YouTube exposes the AI content label as a DOM element with recognizable text or an aria-label, so a script that watches for changes on the page can detect it just like a native app would. This first example does the bare minimum: it hides the cards visible when the page loads.
// weedout-mini.user.js
// Hides, when the page loads, the cards with the AI label
document.querySelectorAll('ytd-rich-item-renderer').forEach((card) => {
const badge = card.querySelector('[aria-label*="IA" i], .badge-style-type-ai-generated');
if (badge) card.style.display = 'none';
});The limitation of this first script is obvious: it only runs once, when the page loads. YouTube keeps injecting new cards constantly through infinite scroll, so you need a MutationObserver for the filter to keep working as you browse. This second version, built as a Tampermonkey userscript, adds exactly that and replicates Weedout’s dim mode:
// ==UserScript==
// @name Homemade Weedout
// @match https://www.youtube.com/*
// @grant none
// ==/UserScript==
(function () {
const AI_SELECTOR = '[aria-label*="IA" i], .badge-style-type-ai-generated';
const DIM_MODE = true;
function weed(node) {
const card = node.closest('ytd-rich-item-renderer, ytd-video-renderer, ytd-reel-item-renderer');
if (!card) return;
if (DIM_MODE) {
card.style.opacity = '0.15';
card.style.pointerEvents = 'none';
} else {
card.remove();
}
}
const observer = new MutationObserver(() => {
document.querySelectorAll(AI_SELECTOR).forEach(weed);
});
observer.observe(document.body, { childList: true, subtree: true });
})();Installed with Tampermonkey (it works the same on Windows, macOS, and Linux, in any Chromium-based browser or Firefox), this script runs the same check every time the observer detects a change in the DOM, no matter how much you scroll. To confirm it’s active, open the DevTools console and count how many nodes ended up dimmed with document.querySelectorAll('[style*="opacity: 0.15"]').length: if the number rises as you scroll, the filter is working.
💡 Tip: always start in dim mode, not remove mode. That way you can visually confirm that the detected label actually corresponds to a real AI video before trusting that the filter isn’t making mistakes.
How to start trying it out
If you’re running macOS 13 or later, installing the real app is straightforward: open the Mac App Store, search for “Weedout”, pay the one-time $1.99 price, and enable whichever mode you prefer (dim or remove) from its settings. You can also optionally enable automatically skipping AI-labeled Shorts in the player, as described on the product’s official page.
If you’re on Windows or Linux, there’s no native version: Weedout depends on macOS frameworks and is distributed only through the Mac App Store. The closest alternative today is installing a userscript manager like Tampermonkey (available for Chrome, Edge, and Firefox on Windows, Linux, and macOS) and loading a script like the one above, adjusting it to the actual selector YouTube uses in your region and language.
In both cases, the recommendation is the same: start in dim mode for at least one day of normal use, manually check some of the dimmed videos to confirm they actually carry the AI label, and only then switch to remove mode if the filter has convinced you.
Impact and analysis
Weedout’s real contribution isn’t technical, it’s a product decision: it turns a passive YouTube label into an active user action. Until now, AI content disclosure existed, but it changed nothing about the experience of browsing the feed. That’s exactly what the app solves.
The honest limitation lies in the very term the developer uses for undetected content: “slop”. If a channel doesn’t mark its video as AI-generated (whether by oversight or on purpose), Weedout has no way to detect it because it doesn’t guess or run its own classifier. This is a deliberate design decision, not a bug: the developer prefers zero false positives at the cost of letting undeclared content through.
⚠️ Heads up: Weedout isn’t an AI detector. It’s a filter for a label that YouTube itself already calculates. If an AI video isn’t marked, the app won’t touch it.
As for privacy, the model is easy to audit: it doesn’t ask for login, doesn’t create an account, and, according to its documentation, all processing runs on the user’s Mac. That sets it apart from content moderation tools that depend on an external server to analyze each video, with the latency and privacy cost that entails.
What’s next
The developer hasn’t published a public roadmap or date commitments for bringing Weedout to Windows, Linux, or a browser as a cross-platform extension. For now, the declared scope is macOS 13 and up, distributed exclusively through the Mac App Store.
What does remain open is the reporting channel: any user can write to [email protected] with the link to a video that should have been filtered but wasn’t, which suggests the detection criteria can be adjusted over time as YouTube changes the visual markup of the “Made with AI” label.
📖 Summary on Telegram: See summary
Try it yourself: if you have a Mac running macOS 13 or later, open the Mac App Store, search for Weedout, and enable dim mode to see how many AI videos you’d been scrolling past without realizing it, right in your own feed.
Frequently Asked Questions
What exactly does Weedout do?
Depending on the mode chosen, it hides or dims videos that YouTube already labels as “Made with AI” in the feed, search, related videos, playlists, and Shorts.
Does Weedout detect AI on its own?
No. It relies exclusively on the disclosure label YouTube already applies to the video. It doesn’t use its own heuristics or an additional classification model.
How much does it cost and where can I get it?
It costs $1.99, a one-time payment with no subscription, and it’s distributed only on the Mac App Store for macOS 13 or later.
Is there a version for Windows or Linux?
There’s no official native version. The alternative available today is a homemade Tampermonkey userscript that replicates the same filtering logic on any operating system.
Does Weedout store or send my browsing data?
According to its documentation, it doesn’t create accounts, doesn’t track the user, and all processing runs locally on the Mac, without sending data to an external server.
What happens if an AI video slips through undetected?
The developer asks that you report it by email to [email protected], including the video’s link, so the case can be reviewed.
References
- Weedout for YouTube: the product’s official page, with pricing, requirements, and detection scope.
- YouTube Help Center: the altered or synthetic content disclosure policy that underlies the “Made with AI” label.
- Wikipedia: Synthetic media: general context on content generated or altered with artificial intelligence.
- Tampermonkey: the cross-platform userscript manager used in the homemade filtering example.
📱 Enjoying this content? Follow @programacion on Telegram for daily tech content in Spanish: quick summaries, fresh content every day.
0 Comments