Four Simple Steps to Fix Bad YouTube Recommendations
YouTube's recommendation algorithm can go off track when it lacks clear signals from viewers, but a handful of settings changes can retrain it. Adjusting watch history, feedback and account settings can restore more relevant suggestions.
YouTube hosts such a vast library of videos that most users rely on its recommendation system rather than searching manually. The algorithm draws on viewing signals to predict what people want to watch, but it can occasionally misfire, filling the homepage, sidebar and end-of-video suggestions with irrelevant content.
Enable watch history
The first step is confirming that watch history tracking is turned on through the Google Account's Data & Privacy settings, under Personalization. Without it, YouTube has little to learn from and may even show a blank homepage. The same menu allows users to review auto-delete settings, which can clear history every three, 18 or 36 months.
Send clear positive and negative signals
The most impactful step is giving YouTube explicit feedback. Watching liked videos in full and clicking Like, subscribing to favorite creators, and leaving comments all reinforce good matches. Negative feedback matters too — hitting Dislike, choosing "Not interested," or selecting "Don't recommend channel" helps filter out unwanted content. Unsubscribing from unused channels also helps keep the profile focused.
Limit accidental viewing
Shared devices, such as a family TV or tablet used by children, can distort recommendations with unrelated watch activity. Setting up separate accounts, locking an account with a passcode, or using incognito/private browsing while watching unrelated content can prevent this kind of interference.
Manage history selectively
The Google Account's history management tool lets users review, filter and delete individual entries by date or search term. While wiping the entire history can help when recommendations are far off track, a more targeted approach — clearing only a specific time range while preserving the rest — often strikes a better balance for retraining the algorithm over time.
