Bouncer: Browser Extension Filtering Generative AI Content from Social Feeds

An introduction to Bouncer, a browser extension that detects and removes generative AI images and unwanted feed content using an in-browser lightweight model an

tau · October 7, 2026

#Bouncer #BrowserExtension #AIContentFilter #GenAIBlocker

Bouncer: Browser Extension Filtering Generative AI Content from Social Feeds

As user fatigue mounts over the overwhelming flood of generative AI (GenAI) imagery and algorithmic clutter across social media timelines, a new client-side tool offers a practical solution. Developer Millan Philipose (@fakemillanp) has released Bouncer, a browser plugin designed to automatically detect and remove unwanted generative AI content from social feeds.

Conceptual illustration of the browser extension Bouncer automatically identifying and filtering generative AI images and unwanted feed content using natural language prompts and an in-browser lightweight model

Image source: @fakemillanp

Standard moderation tools native to social platforms have long been limited to rigid keyword blacklists or blunt account-level muting, leaving users largely defenseless against synthetically generated images or repetitive promotional patterns. Operating directly inside the user's browser, Bouncer dynamically inspects feed elements and removes content matching user-defined criteria.

In-Browser Lightweight Model and Client-Side Detection

A central technical design choice in Bouncer is its reliance on an in-house trained, lightweight AI model ("small model") executing locally within the browser, rather than tethering every feed scan to external cloud APIs.

Relying on commercial vision APIs introduces prohibitive latency, recurring API costs, and significant privacy concerns, as users' browsing histories and viewports would need to be transmitted to remote servers. By embedding a compact neural model directly into the extension runtime, Bouncer evaluates whether images displayed in the timeline are synthetic AI creations directly on the user's machine.

  • Local Inference: Eliminates continuous remote API calls, evaluating images in-browser with minimal network overhead.
  • Real-Time Feed Cleansing: Dynamically intercepts and deletes synthetic GenAI images as users scroll through their social timelines.

Natural Language Filtering: Removing Anything Described in Words

Beyond detecting standard AI-generated imagery, Bouncer provides a flexible natural language rule interface that allows users to customize what gets filtered out.

In response to community questions, developer Millan Philipose clarified that "Bouncer can filter anything you can describe in words." This extends filtering capabilities well beyond raw image classification into contextual and topic-based social moderation.

  • Promotional and Course Spam: Users can instruct the extension to eliminate commercial solicitations, such as aggressive SEO and marketing course pitches.
  • Contentious Discussions: Feeds can be stripped of repetitive online arguments, including debates over AI adoption.
  • Context-Aware Rules: Replaces simplistic single-keyword blocking with nuanced natural language descriptors tailored to individual preferences.

Free Availability, Monetization Roadmap, and Practical Caveats

Bouncer is currently available as a browser plugin with its core filtering features offered for free.

Prospective users should keep several practical caveats in mind:

  1. Future Paid Features: While the current core feature set remains free, the developer has stated that future advanced capabilities will likely transition into paid tiers.
  2. Detection Boundaries: Because the lightweight model is constrained by client resources, DOM structural updates or subtly generated images may occasionally cause false positives or missed detections.

For users seeking to reclaim visual clarity on their timelines, Bouncer stands out as a practical utility.

Sources