A video app has a huge number of videos to pick from. How does it choose which ones to show you? Part of the answer is a recommendation system. A recommendation system picks videos for you from a huge number of videos. Here is one design for such a system. Candidate generation is the first stage of recommendation: given a query, the system generates a set of relevant candidates. In one common recommendation design, the system starts from a potentially huge collection and generates a much smaller subset of candidates. Here is how that can work, one step at a time, with an example from Google's course notes.

Content-based recommending uses similarity between items to recommend items similar to what the user likes. Google's course notes give an example. If user A watches two cute cat videos, the system can recommend cute animal videos to that user. A recommendation system suggests videos similar to ones a person watched. Two more stages are described in the course notes. In a scoring stage, the system can use a more precise model relying on additional queries. In a re-ranking stage, the system can remove items the user explicitly disliked or boost the score of fresher content. Re-ranking can also help ensure diversity, freshness and fairness.

Next, think about a feed of posts. To a first approximation, generating a user's feed is a matter of ranking all the posts that can be shown. For every post in a user's feed, the algorithm receives feedback about whether and how the user engaged with it. Recommendation algorithms are largely limited to looking for patterns in behavioral data. Now look at one real example. It comes with dates. In a September 2021 post, YouTube said its system sorts through billions of videos to recommend content tailored to a viewer's specific interests. In 2021 YouTube said its system learns every day from over 80 billion pieces of information it calls signals. YouTube said its system does not follow a set formula but develops dynamically as a viewer's habits change.

The clues have changed over time. As of September 2021, YouTube said it learned in 2011 that clicking on a video does not mean the viewer actually watched it. As of September 2021, YouTube added watch time to its recommendations in 2012. As of September 2021, YouTube said that when it first added watch time to recommendations it saw an immediate 20% drop in views. As of November 2026, YouTube's recommendations take into account how long people watch. Take a moment to read the dates with care, and look at each year, since they tell you when each thing was said. Each date is the one that goes with the fact before it. Take your time.

Some experts say that a system like this is built to keep people watching. Here is what they say. Here is when they said it. As of August 2026, Common Sense Media says the YouTube algorithm isn't designed with kids in mind and is designed to keep users engaged. As of August 2026, it says Up Next and autoplay encourage hours of uninterrupted viewing, often leading kids to lose track of time. As of November 2026, it says YouTube's business model rewards watch time: the longer users watch, the more YouTube makes in ad sales. As of November 2026, KidsHealth says most apps, games and streaming platforms are designed to keep users involved as long as possible.

There are ways to stop, and they are good to know about. Here is a short list of what a viewer can do, and of what is said about it. As of November 2026, if Autoplay is on and a viewer wants to stop the next video, they can tap Cancel at the end of a video. As of November 2026, on the Time management tab, a viewer can change the Take a break reminder and Auto-play settings. As of November 2026, YouTube says the time watched profile lets a viewer check their daily average watch time. As of November 2026, KidsHealth says to turn off autoplay on streaming platforms. Each of these is something a viewer can look for in the settings, and it is good to know where to find it. Read each fact with its date.

Viewers can also steer the picks. As of November 2026, on certain pages, such as Home and Watch Next, a viewer can mark a video as Not interested. As of November 2026, YouTube says Not interested and Don't recommend channel feedback may be used to tune recommendations. As of November 2026, YouTube says that if a viewer deletes watch history, it won't base future video recommendations on that content. As of November 2026, disliking a video is a signal that can change what YouTube recommends. Read each of these with its date, and think about which of them you might want to try yourself on a video app that you use.

Age rules apply as well. As of November 2026, Common Sense Media says YouTube requires users to be at least 13 years old to create an account. As of November 2026, YouTube's terms of service say children of all ages may use YouTube and YouTube Kids if a parent or guardian enables it. As of November 2026, the COPPA Rule was designed to protect children under age 13. As of November 2026, websites covered by COPPA must notify parents and get their approval before collecting a child's personal information. Read each fact with its date, and think about which ones apply to you. They are short, so it is easy to read them twice.

As of November 2026, YouTube Kids says its Recommended videos are selected by its algorithm without human review. Here is what that looks like for a child who uses the app. Think about who picks what the child sees. In this case, the answer has more than one part, and each one is a fact that you can look up. As of November 2026, in YouTube Kids, recommended videos are also based on the content setting a parent selected. As of November 2026, if a child's profile is set to Approved content only, Recommended videos are drawn from content the parent has approved. As of November 2026, a parent can stop new video views from being used as signals to recommend other videos.

Words to know

• candidates: possible picks
• algorithm: selection program
• diversity: a mix
• engaged: kept watching