YouTube's recommendation system aims to identify the most relevant content for each user at any given moment. Personalised recommendations are designed to help viewers easily find content that they'll enjoy, from exploring new topics and creators to staying connected with their favourites.
The algorithm has two goals
- To help each viewer find videos that they want to watch
- To maximise long-term viewer satisfaction
The system analyses each viewer's profile in real time, which considers device, time of day and past habits to surface and rank the most relevant content, leading to a highly personalised and unique viewing experience for everyone.
The algorithm follows the audience
Creators often focus on 'the algorithm', but what matters is what viewers enjoy. Instead of asking 'Does the algorithm like my content?', creators should focus on their 'audience' and ask, 'Does my audience like this?'
Ultimately, YouTube's recommendations are driven by what viewers watch and enjoy. The algorithm is learning from audience signals. These signals can be categorised into:
- Viewer personalisation: This takes into account the signals that help us understand a user's preferences.
- Content performance: This is all about how well the video performs when it's offered to viewers, like whether they choose to click, watch or positively engage with the content.
Beyond personalisation and performance, external factors also influence YouTube's recommendations.
Recommendations are personalised
We refer to the signals that help us understand a user's preferences as 'personalisation'. Think of YouTube's recommendation system like suggesting a book or a movie to a friend. It understands your audience's unique tastes, past likes and dislikes, even what similar users enjoy. This is essentially automated word of mouth on a massive scale, designed to truly satisfy a viewer's preferences. It achieves this by primarily examining two signal types: a viewer's watch history (content that they've enjoyed) and their interest affinity (explicit and inferred topics that they're into).
- Watch history: Our systems aim to determine content that viewers might enjoy by learning from their watch history. This includes what videos they choose to watch, ignore or dismiss ('not interested' feedback), how long and how much of a video they watch and other viewing routines. We also consider content that they've searched for or engaged with, such as likes, shares, comments, 'not interested' feedback and survey responses. Additionally, we factor in channels that a user subscribes to and the languages of content that they consume.
- Interest affinity: Our systems learn about viewer's preferences by recognising a viewer's favoured themes, topics and formats, as well as what's popular among similar audiences. It also learns from what videos similar viewers also enjoyed, like 'viewers of video A also watched video B'.
What does this mean for your channel growth strategy?
- Sustainability is key: A sustained presence helps your audience embrace your channel as a routine that they anticipate, keeping it top of mind. Being there for your community means regularly delivering valuable content that they can rely on and crafting a recognisable experience. Plus, a consistent title and thumbnail style makes your videos instantly identifiable, helping viewers quickly choose what to watch.
- It is essential to get to know your audience: Unlocking greater discoverability starts with understanding your audience's interests. The Audience tab in Analytics provides powerful insights, showing you preferred formats and what 'channels and content your audience is also watching'. This helps you spot market shifts and adapt your strategy.
- Don't overthink your publish time: The algorithm aims to deliver the right content to viewers whenever they visit YouTube, regardless of upload time. While publishing when your audience is most active might lead to more immediate views, we haven't observed any evidence that it affects long-term viewership. This is to say that publish time is not known to impact a video's long-term performance. Note, however, that if you're scheduling a Premiere or going live, consider your audience's active times to maximise engagement.
- Help viewers discover more of your content: Make it easy for viewers to find and watch more of your content. For example, you can develop content series to encourage continued viewing, and use clear calls to action (example: If you liked this, then watch…), playlists and end screens to surface more content. Remember, positive viewer interactions are a strong signal for the recommendation system.
- Quality over quantity: Prioritise consistent quality content over a high frequency of uploads. This approach fosters sustained audience engagement, which is key for growth. Remember, true consistency also involves making it sustainable for yourself. Taking breaks is a healthy and encouraged part of a long-term creative journey – the algorithm doesn't penalise creators for doing so! If you need some time off, letting your audience know can help you re-engage them smoothly upon your return.
- Experiment to adapt: Your audience's interests will naturally evolve – just think how much your own tastes have changed over time! Our system adapts to these shifts. Experimenting with new content is a great way to stay connected with your audience. The algorithm uses fresh performance data for each individual video, rather than relying on past results, so a single experiment that doesn't land won't necessarily hinder your channel's potential. Remember that YouTube aims to give as many videos as possible a chance to be discovered by audiences who are interested in them! It's worth bearing in mind, though, that frequently sharing content that doesn't resonate can, over time, affect your channel's overall performance.