YouTube plans to stop showing users recommended videos in some places when watch history is turned off entirely, including the homepage
Recommended videos keep users engaged on YouTube, but starting today, YouTube will stop showing recommended videos in some places, such as the homepage …
Context & Ripple Effects
YouTube had already added user-level controls such as blocking recommendations from selected channels and choosing topics, while its Kids product was being positioned with an option to turn off algorithmic suggestions altogether. This change extends that control from individual recommendation inputs to the underlying watch-history signal.
The move matters because the homepage is a primary discovery surface: users who fully withhold viewing history will see a less personalized YouTube experience rather than a substitute recommendation feed.
First-order effects
- Users who turn watch history off entirely lose recommended videos in designated surfaces, including the homepage.
- YouTube must distinguish history-disabled sessions from ordinary personalized viewing flows and present a reduced-discovery interface to those users.
Second-order effects
- Creators and channels that rely on homepage discovery may receive less exposure from the subset of viewers who disable history; subscriptions and direct searches become relatively more important paths to viewing.
- The policy makes YouTube's existing preference tools more consequential: channel and topic controls preserve some personalization, while fully disabling history now carries a clearer trade-off in discovery.
Third-order effects
- If this pattern persists, recommendation engines increasingly become conditional on users supplying behavioral signals, with explicit controls defining how much algorithmic distribution a platform can provide.
- The change also fits a broader shift toward constraining recommendation systems in sensitive contexts, reflected in YouTube's later limits on repeated sensitive-topic recommendations for US teens.
The trend: Large platforms are making algorithmic discovery more configurable, trading some engagement potential for clearer user control over the data that powers recommendations.