Open-source project
Stevoisiak/Stevos-AI-Blocklist avatar
Stevoisiak/Stevos-AI-Blocklist

Stevo's AI Blocklist: an adblock filter list that hides generative AI features

A filter list that hides website features which use Generative AI & content labeled as AI Generated.

1,016 stars24 forksAdblock Filter ListMIT

At a glance

What is it?
Stevo's AI Blocklist is a text filter list for uBlock Origin and AdGuard that hides specific site features built on generative AI, along with content that the site itself labels as AI generated, and it ships as two plain files.
Who is it for?
Take Stevo's AI Blocklist if your objection is to generated answers and AI labelled media rather than to advertising, and if you are willing to subscribe to a list that someone else maintains and that will change under you as sites move their markup. Leave it out if you need the features it removes, because the filters are matched by URL pattern and site layouts change without notice.
Can I use it commercially?
Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?
Yes. The repository received new commits within the last day.
What is it written in?
Mainly Adblock Filter List, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on October 23, 2026, and from our analysis. They are not legal advice.

Editorial analysis

The target is generated content, not advertising

This is a filter list in the adblock sense, hosted as a plain text file that a blocker downloads, but what it filters is not advertising. Its stated purpose is to hide specific website features that use generative AI, and content that a site labels as AI generated. The list is built for uBlock Origin and AdGuard, and the documentation walks through installing it on desktop Linux and PC, on macOS, on Android and on iOS and iPadOS, with per browser instructions for Firefox, Chrome, Microsoft Edge, Brave and Safari. That portability comes from the fact that the artefact is only a list of filter rules. The list is published from a branch of the repository and fetched over a raw file URL, so the same string works in every one of those setups.

The main list covers named features on named sites

The documented examples make the scope concrete. Google's AI Overviews and AI Mode are on the list. On YouTube the covered items are the Ask button, video summaries, auto dubbing, automatic chapters and the Super Resolution upscaling. Copilot buttons are covered on GitHub, Bing, Microsoft 365 and the Azure Portal. The list also reaches into media sites, hiding Pixiv and DeviantArt images that are labelled as AI generated. Commerce is included through Amazon Rufus and review summaries. Reddit Answers and recommended posts from AI subreddits are covered, along with Facebook's AI chat and posts carrying an AI info label, TikTok videos tagged as AI generated, and on X the Grok buttons, posts from Grok and posts labelled as made with AI. Nothing in the list is a site-wide block, which is what keeps it usable rather than blunt.

Firefox desktop takes a subscribe link or a pasted URL

On desktop Firefox the shorter route is to install uBlock Origin, left click the subscribe link in the documentation and press Subscribe to import the filter list. The documentation also explains the manual fallback for people with more than one blocker installed, which is a real conflict on Firefox. In the uBlock dashboard, the Filter lists tab has an Import section where the raw URL is pasted.

text
https://raw.githubusercontent.com/Stevoisiak/Stevos-AI-Blocklist/refs/heads/main/GenAI-Blocklist.txt

That URL is the same one used by every other platform in the documentation, and it points at the file in the repository rather than at a subscription service. Applying changes finishes the import. On Brave desktop the same subscribe link works, but only after uBlock Origin is enabled in the browser settings, since Brave ships with Manifest V2 extensions switched off by default.

Chrome and Edge need user scripts allowed first

The Chrome and Microsoft Edge route uses the AdGuard browser extension from the Chrome Web Store, and it has one prerequisite that catches people out. In the AdGuard settings, on the Filters tab under Custom, the extension settings link has to be opened and Allow User Scripts has to be enabled. Only after that does the subscribe link register the list properly, and the reader is then asked to confirm the custom filter with Next, optionally marking it Trusted, and finally to click Add. On Android Firefox the same import path exists in the uBlock Origin extension settings, under the Filter lists tab, scrolled to the Import section, followed by Apply Changes. The pattern across browsers is consistent: the list is added once and updates itself from then on.

Apple platforms route the list through AdGuard's own settings

On macOS, Safari is handled by AdGuard Mini, where you accept the agreement, enable the seven AdGuard extension checkboxes in Safari settings and allow the extension on every website, then paste the same URL into Settings, Filters, Custom filters and switch the custom filters toggle from Disabled to Enabled. On iOS and iPadOS, the AdGuard app is installed from the App Store, each AdGuard extension is enabled under Settings, Apps, Safari, Extensions, and the list is turned on under Safari Protection, Filters, Annoyances, where the documentation insists on tapping the text of the category rather than its on and off icon. Brave on iOS takes a different route through Shields and Privacy, Content Filtering, Add Filter by URL. Each platform needs a different number of taps for the same file.

The extra list is more coverage with a worse failure mode

A second file, GenAI-Blocklist-Extra.txt, is offered as an optional addition. The documentation is direct about the trade: those filters are more subjective or experimental, they are more prone to accidentally blocking non-AI content, and they can break site functionality. The examples given are an AI category in headers on news sites, customer support chatbots that have to be used before a human agent can be reached, and a YouTube rule that removes auto dubbing when a video is opened directly by URL but causes a visible page refresh even on videos with no auto dubbing. A Zoom entry about highlights and automatic chapters is listed as well. Read that list as a set of policies you opt into one by one rather than as an automatic companion to the main file.

MIT terms, plain text files, no tagged releases

The repository holds very little: a gitattributes file, a GitHub directory, a CONTRIBUTING file, the main filter list, the extra filter list, a LICENSE.txt file and the README. The licence is MIT, so the rules themselves can be reused and modified under the terms of that licence. The substance of the project is therefore one text file that a browser extension fetches and applies, and the maintenance signal is the push history rather than a release cadence: there are no GitHub releases at all, and the last push to the repository was on 2026-10-15. A list keyed to markup that Google, YouTube, Reddit and X change whenever they feel like it has no way to be versioned safely, so the file is simply replaced in place.

Editorial conclusion

Take Stevo's AI Blocklist if your objection is to generated answers and AI labelled media rather than to advertising, and if you are willing to subscribe to a list that someone else maintains and that will change under you as sites move their markup. Leave it out if you need the features it removes, because the filters are matched by URL pattern and site layouts change without notice. Before you subscribe, read the extra list description and decide whether the risk of blocking customer support chatbots or breaking video pages is worth the coverage.

Frequently asked questions

Which browsers does Stevo's AI Blocklist support?

It works with uBlock Origin and AdGuard, and the documentation covers desktop Firefox, Chrome, Microsoft Edge and Brave on PC and Linux, macOS Safari, Android Firefox, and Safari and Brave on iOS and iPadOS.

What does Stevo's AI Blocklist actually hide?

It hides named generative AI features on specific sites, such as Google AI Overviews and AI Mode, the YouTube Ask button, video summaries, auto dubbing and automatic chapters, Copilot buttons on GitHub and Bing, Amazon Rufus, and posts or images that a site itself labels as AI generated.

Why does Chrome need Allow User Scripts before the list works?

The AdGuard extension route requires that option to be switched on in its extension settings, under the Filters tab and Custom, before the subscribe link can register the custom filter list.

Is the extra filter list safe to enable?

The documentation describes those filters as more subjective or experimental, prone to blocking content that is not AI related and prone to breaking site functionality, with customer support chatbots and a YouTube rule that forces a page refresh given as examples.

Official sources

  1. Issues
  2. License: MIT
  3. README
  4. Stevoisiak/Stevos-AI-Blocklist on GitHub
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/stevoisiak-stevos-ai-blocklist.svg)](https://hysenlabs.com/projects/stevoisiak-stevos-ai-blocklist)

Was this useful?