August 29, 2026
YouTube's stated goal for its recommendation system is maximizing total time spent watching across the platform, not maximizing any single video's view count. That's why average view duration and session time (does a viewer stay on YouTube after this video, or leave) weigh so heavily — a video that gets a lot of clicks but loses viewers immediately signals a bad recommendation, even with a high view count.
Thumbnail and title click-through rate matters — a video nobody clicks can't perform — but YouTube also tracks whether people who clicked actually stayed. A high CTR paired with a low average view duration reads as a misleading thumbnail or title, and tends to suppress future recommendations rather than help them, which is why pure clickbait strategies usually stop working after a channel's first few videos.
YouTube has shifted heavily toward recommending videos based on topic and viewing behavior rather than subscription status — most watch time on the platform now comes from recommendations to non-subscribers, not from subscriber notifications. A channel's subscriber count still signals some baseline trust, but a channel that consistently produces videos people finish watching will get recommended to non-subscribers regardless of subscriber count.
Strong retention in the first 15-30 seconds predicts whether a video gets tested further, similar to TikTok's early-window logic. Session time — whether viewers click into another video on the channel afterward — is a growing signal, which is part of why playlists and end-screen suggestions to related videos on the same channel measurably help. And consistent upload topics help YouTube's system understand who to recommend a channel to, more than consistent upload schedule alone.