An Overview of YouTube’s Recommendation System

YouTube was created in 2005 as a video-sharing platform. After a year, Google bought YouTube and made it the largest video-sharing platform globally. It was also 2nd in rank as most visited in the world. Of course, its parent company, Google, ranks 1st.  Today, YouTube connects more than 2 billion people with 720,000 hours of videos uploaded daily. 

According to a video marketing agency Hong Kong, YouTube has been continuously working on improving its features and tools. The platform also aims to help brands and creators get an optimal response on every video that they upload. Let us now take a deeper look at YouTube’s recommendation system.

What is YouTube’s Recommendation System?

YouTube believes that there is a unique audience for every video. So, it is their job to find the best audience for an uploaded YouTube clip. The platform has spent more than a decade building the best experience among YouTubers. So, it comes up with a recommendation system to boost overall viewership among channels, subscribers, and video views. YouTube’s recommendation system started in 2008. During that time, the system ranks popular videos to create a big trend on the page. Today, it is entirely different. The recommendation system is now base on a YouTuber’s viewing habits and online activities. They are tailored according to specific interests coming from billions of video content on the platform. 

A digital marketing speaker Hong Kong defines YouTube’s recommendation system as helping people find videos that they are interested in and get value out of them. YouTube’s recommendation system works in two main areas – YouTube’s homepage and the “Up Next” panel. YouTube’s homepage is what people see first as they open their accounts. Here, YouTube’s recommendation system displays personalized suggestions of the latest news and trends. YouTube’s “Up Next” panel appears after a viewer finishes watching a video. It suggests new videos relevant to the content of what a viewer has recently watched. It also suggests videos that the system thinks the user might be interested in. 

How Do YouTube Personalizes Its Recommendation System?

According to a social media agency Hong Kong, YouTube is the largest social media network for video-sharing. As such, its recommendation system constantly evolves. Its custom curation continuously learns from 80 billion signals. These signals include clicks on a video, reactions from viewers, survey responses, and watch time. 

  • Clicks. Every click on a video is a strong indication of interest from the viewer. After all, nobody will click on something that he or she does not want to watch. Although, clicks on videos do not actually mean watching them. Sometimes, the click only shows an intent to search for more related videos. That is why YouTube enabled the “Up Next” panel. It is to make it easier for viewers to find relevant content that they might be searching for. 
  • Reaction from Viewers. There are different ways in how a viewer can react to a YouTube video. He or she can either comment, dislike, like, or share the video. He or she can also save the video on a playlist or subscribe to the YouTube channel that created it. YouTube’s recommendation system uses these signals to rate a video as positive or negative. Likes and shares as signs of a positive engagement. While dislikes signal negative responses on the part of the viewer. These reactions from YouTube viewers can affect the ranking or star system of a video. 
  • Survey Responses. YouTube also conducts surveys from time to time. It is to make sure if viewers are really satisfied with the videos they are watching. Here, the platform asks participants to rate the value of a video from one to five stars. Such metrics determine how satisfying a YouTube video is. Furthermore, YouTube asks viewers who have given a rating of one or two stars the reason why they rated the video as low in value. Only videos with four or five stars are rated as high-value content. YouTube is aware that not all viewers will respond to the survey. So, its machine learning has been designed to predict potential survey responses to all viewers. 
  • Watch Time. YouTube watch time is the length of time a viewer stayed to watch video content. Watch time allows YouTube to better know what a viewer will most likely watch. For example, most Wimbledon fans watch 20 minutes of highlight game clips and only a few seconds of match analysis. YouTube’s recommendation system will assume that Wimbledon game highlights are more valuable than match analysis.


YouTube’s recommendation system ranks videos by incorporating all the signals mentioned above.  Its main goal is to give viewers the content they love and help creators gain optimal reach on their video content. At the same time, YouTube’s recommendation system also aims to prevent the spread of harmful misinformation.