When a viewer opens YouTube, the platform doesn't just show a random assortment of videos. It presents a highly personalized feed, curated by a sophisticated recommendation system. This system is responsible for over 70% of the time users spend watching videos. Understanding how it works is key to growing an audience.

The process isn't a single, monolithic algorithm. It's a two-stage system designed to first find a broad set of relevant videos and then rank them to find the perfect one for that specific moment.

The first step is called candidate generation. The goal here is to quickly create a large pool of potential videos—hundreds of them—that could be a good match for the viewer. This process relies heavily on a user's past activity and the behavior of similar users.

YouTube analyzes a viewer's entire watch history, including videos they liked, disliked, or commented on. It also looks at their channel subscriptions. But a crucial piece of this stage is co-watch probability, also known as collaborative filtering. The system identifies users with similar viewing habits and then recommends videos that those

This creates a broad but still relevant list of videos. It's a mix of videos directly related to what the user has watched before and videos that are popular among a similar taste profile. This large list is then passed to the next stage for fine-tuning.

The second step, ranking, is where the algorithm gets much more precise. It takes the hundreds of videos from the candidate pool and scores each one based on dozens of signals. The goal is to predict which video the viewer is most likely to watch and enjoy right now.

Key signals for ranking include:

• Watch Time & Session Duration: How long do people watch your video? Does it lead them to watch more videos afterward? The algorithm rewards content that keeps viewers on the platform.
• Engagement: Likes, dislikes, comments, and shares are all taken into account. They are treated as strong indicators of a viewer's satisfaction.
• Click-Through Rate (CTR): When your video is shown on a homepage, what percentage of people click to watch it? This measures how compelling your title and thumbnail are.

While these signals are important for any video, the algorithm has become increasingly focused on watch time from people who are not subscribed to your channel.

For sensitive topics, especially those categorized as "Your Money or Your Life" (YMYL) like news, finance, and health, the algorithm applies an additional layer of scrutiny. In these areas, it prioritizes signals of reliability and authority. This means content from established news organizations, licensed professionals, and reputable institutions is often ranked higher to prevent the spread of misinformation.

Understanding this two-stage process helps creators optimize for the two most important discovery surfaces: the Homepage and the "Watch Next" feed.

• Homepage: This is driven by a mix of personalization and diversification. The algorithm wants to show viewers content from channels they love but also introduce them to new creators based on their broader interests. Success here often depends on creating content with wide appeal that generates strong non-subscriber watch time.

• Watch Next: This feed is highly contextual. The algorithm's main goal is to find the most logical next video for the viewer to watch based on what they just finished. It heavily weighs the topical relationship between videos. Creating series or playlists that encourage sequential viewing is a powerful strategy here.

Ultimately, the recommendation system is designed to create satisfying long-term user journeys. It's not about tricking an algorithm with keywords; it's about creating content that genuinely engages an audience and keeps them coming back for more.

Ready to test your knowledge?

By focusing on viewer satisfaction and understanding the signals that matter, you can align your content strategy with the goals of the YouTube algorithm.