How do recommendation systems (like YouTube or Spotify) decide what to show me?

Recommendation systems are AI tools that try to predict one thing: what you are most likely to watch, listen to or click next. They do this by learning from your behaviour, the content itself and millions of other users.

1. They learn from what you do

The system collects two kinds of signals:

• Explicit signals: likes, dislikes, follows, subscriptions, saving a song to a playlist.
• Implicit signals: what you click, how long you watch, what you skip, what you replay and what you search for.

Implicit signals matter most because there are so many of them. Watching a video to the end tells the system far more than just clicking on it.

2. They find people like you

If you and thousands of other users enjoy the same five artists, and many of them also love a sixth artist you have never heard, the system will suggest that sixth artist to you. This is called collaborative filtering: people who liked what you liked also liked this.

3. They study the content itself

The system also analyses each item, such as a song's tempo and mood, or a video's title, topic and description. This is called content-based filtering. It helps the system recommend brand-new content that few people have seen yet.

4. They narrow down, then rank

With billions of videos and millions of songs, the system works in stages:

1. Pick candidates: quickly select a few hundred possible options.
2. Rank them: score each one on how likely you are to engage with it.
3. Adjust the list: add variety, remove things you have already seen and filter out content that breaks platform rules.

5. They are built to reach a goal

Each platform decides what "success" means, such as more watch time or more return visits. This shapes what you see. A system designed to keep you engaged may show you content that grabs attention, which is not always the same as content that is accurate, balanced or good for you.

Judging what you see

Your feed is not a neutral picture of what is popular or true. It is a prediction made for you. To use recommendations wisely:

• Ask why: "Why might this be recommended to me?" Often it is because it is similar to what you already watched.
• Watch for filter bubbles: if your feed only shows one viewpoint or one type of content, you may be missing the bigger picture.
• Take control: use "Not interested", clear items from your watch history and search for new topics on purpose.
• Check before you trust: a recommended video is not automatically reliable. Verify important information with trusted sources.

Key takeaway: Recommendation systems learn from your behaviour to predict what will keep you engaged. Understanding how they work helps you decide what to watch, rather than letting the algorithm decide for you.