How YouTube Recommends Videos (Inside the Recommendation Engine)
A plain-English breakdown of how YouTube's recommendation system decides which videos to surface, and how to influence it.
YouTube's recommender is two systems stacked: a candidate generator that picks 100 possible videos, and a ranker that orders them for a specific user. Understanding both is how top channels manufacture 'suggested' traffic.
Candidate generation
The system starts with your watch history, subscribed channels and similar-viewer behaviour. Videos sharing metadata, transcript topics or audience overlap with your recent watches enter the candidate pool.
Personalized ranking
The ranker scores each candidate for predicted watch time. Videos with higher session-length value to a specific viewer float to the top of the Home feed and Suggested rail.
Influence the system
Series-format videos, tight audience overlap with your niche and prompt cross-linking in end screens all raise a video's ranker score. Real early views amplify the signal.
Ready to accelerate your growth?
Grow your YouTube channel with real views, subscribers and watch time. Real, drip-fed campaigns from $9. Instant delivery start, safe for monetisation.
Browse YouTube packages →Frequently asked questions
Why does my video stop getting recommended?
Ranker score drops when average view duration falls or user negative feedback rises (dislikes, 'not interested').
Does buying views help recommendations?
Real early views raise the baseline signal the ranker uses, expanding the recommendation pool.
How long does the algorithm take to recommend a new video?
The first 48 hours are decisive — most recommended-traffic videos hit their peak Suggested placement within week one.