YouTube has shared new details about how users can influence what appears on their recommendations page. YouTube explained that the recommendation system relies heavily on viewing activity such as watch history, clicks and engagement patterns. As a result, small changes in how users interact with videos can change what appears on the homepage and suggested feeds.
Tools that shape the algorithm
YouTube pointed to several built-in controls that viewers can use to manage recommendations. One of the most significant is the watch history. Clearing or pausing watch history can change the types of videos that appear in recommendations, since the system uses that history to understand viewing preferences.
Users can also select “Not interested” on a video to signal that they do not want similar content suggested again. Another option, “Don’t recommend channel,” removes videos from a specific creator from appearing in the recommendation feed.
The platform also allows users to reset their recommendation feedback through Google account settings. Doing so clears previous “Not interested” or “Don’t recommend channel” signals. However, the reset applies to all feedback at once rather than individual selections.
These controls are mainly available for users who are signed in to their Google accounts. Logged-out viewers receive fewer personalized recommendations and have less direct control over how the algorithm responds to their activity.
Limited direct controls
Even with these tools, YouTube still offers limited filtering options. Users cannot block specific topics or keywords from appearing in recommendations, and controls across Shorts, autoplay and search are more limited than the main homepage feed.
YouTube has also tested features such as a “Your Custom Feed” option, which is meant to prevent recommendations from shifting too heavily after watching a single video. The feature has been in testing since late 2025 and has not yet been widely released.
The explanation gives viewers a clearer look at how their activity influences recommendations, though the platform has not fully disclosed how the broader algorithm works.