The YouTube algorithm is the set of recommendation systems that decides which videos to show each viewer on the home page, in suggested videos, in search and in the Shorts feed. YouTube says it focuses on viewers rather than videos, using signals such as what people click, how long they watch and how satisfied they say they are.
People talk about the YouTube algorithm as if it were one secret formula. In reality it is a set of recommendation systems that YouTube has explained in public quite a lot, through its help centre and its official blog.
YouTube's help pages and blog describe the signals its recommendation system uses, including:
The important idea is that the system is built around predicting what each viewer will want to watch and feel good about having watched. It is not scoring videos in isolation.
The same principles apply across the home page, suggested videos beside and below the player, the Shorts feed and, to a lesser extent, search. Search leans more on how well a video matches the query, which is where YouTube SEO comes in.
You cannot control the algorithm, but you can control the things it responds to:
When a channel jumps between unrelated topics, each new video has to find its audience from scratch. When a channel has a clear format and audience, the system has more evidence about who to show the next upload to. That is why it usually pays to start a brand channel with a defined format or series rather than one-off videos.
YouTube says its systems aim to match each viewer with videos they are likely to watch and value. Channels that consistently make videos a specific audience chooses to watch tend to be recommended to more people like them. There is no published advantage for size or for brands.
YouTube has not published an ideal upload frequency. A steady schedule you can keep up with, at a quality your audience values, matters more than volume. Fewer, better videos usually beat frequent uploads that viewers skip.
YouTube lists likes, dislikes, shares and feedback such as Not interested among the signals its recommendation system uses. They sit alongside watch time and satisfaction surveys, so they help, but they are not the whole picture.