Show HN: Weedout – Safari extension that hides YouTube AI-labeled videos
A macOS app called "GET WEEDOUT" ($1.99 one-time purchase) filters YouTube videos tagged with YouTube's own "Made with AI" disclosure badge from home feed, search results, related videos, playlists, and Shorts Detection relies entirely on YouTube's official labeling system rather than heuristic or ML-based detection, avoiding false positives on unlabeled content The app offers three modes: full removal, dimming (visual suppression with verification option), and optional auto-skip for AI-labeled
Analysis
TL;DR
- A macOS app called "GET WEEDOUT" ($1.99 one-time purchase) filters YouTube videos tagged with YouTube's own "Made with AI" disclosure badge from home feed, search results, related videos, playlists, and Shorts
- Detection relies entirely on YouTube's official labeling system rather than heuristic or ML-based detection, avoiding false positives on unlabeled content
- The app offers three modes: full removal, dimming (visual suppression with verification option), and optional auto-skip for AI-labeled Shorts
- All processing runs locally on the user's Mac with no accounts, tracking, or data collection required
- On a real feed, AI-labeled content is removed in approximately half a second, with cached content never appearing at all
Why It Matters
This reflects a growing consumer demand for agency over AI-generated content curation, as platforms increasingly populate feeds with AI-generated material. The tool's reliance on YouTube's own disclosure system rather than independent detection highlights the limitations of current AI content identification and the gap between platform self-regulation and user expectations. It also demonstrates the emerging market for privacy-first, local-first content filtering tools as alternatives to cloud-dependent solutions.
Technical Details
- The app operates as a macOS-native client (requires macOS 13+) that intercepts and filters YouTube interface elements before they render to the user
- Detection mechanism is purely rule-based, matching YouTube's official "Made with AI" badge metadata rather than employing computer vision, audio analysis, or ML classification
- Filtering targets multiple YouTube surfaces simultaneously: home feed, search results, related video shelves, playlist queues, and the Shorts player
- The "dim mode" provides a middle-ground approach that visually suppresses AI-labeled content while preserving the ability to review and override the filter
- No network requests to external servers are made for detection; all logic runs client-side, eliminating privacy concerns associated with telemetry or account linking
Industry Insight
- YouTube's "Made with AI" labeling system, while intended as transparency, is being repurposed by third-party tools as a de facto content filter standard, suggesting platforms should anticipate and potentially integrate user-controlled filtering natively
- The one-time purchase, no-tracking model positions this against the dominant subscription-and-data-collection paradigm, indicating a viable niche for privacy-conscious AI tooling
- As AI-generated content volume continues to grow, demand for user-side curation tools will likely increase, pushing platforms to either improve native filtering options or risk driving users toward third-party workarounds
Disclaimer: The above content is generated by AI and is for reference only.