Skip to content

How Does YouTube’s Algorithm Work?

You have probably heard that the YouTube algorithm decides what people watch on YouTube most of the time. It decides 70% of what is watched, and according to Pew Research Center, 81% of American YouTube users watch videos recommended by this algorithm.

That means YouTube algorithm recommendations count a lot for any content creator working on getting more YouTube views or build a brand out of YouTube marketing strategies. The question now is, how do you optimize your channel and video to work with this algorithm, not against it?

To answer this question, we are going to have a look at the history of YouTube’s primacies when it comes to helping viewers discover new videos. We will also tell you how this algorithm works, including the latest changes. 

A brief history of the YouTube algorithm

The first YouTube video was uploaded in 2005. Today, people are uploading over 500 hours of video on YouTube every minute. Not only that, over 4 billion videos are viewed on this platform every day. 

But how do users find what they want to watch? The short answer is that continuous changes in YouTube’s algorithm make that possible. However, here is the long answer:

2005-2012: View count (a.k.a. clicks)

Between 2005 and 2012, YouTube rewarded videos that got clicks, rather than those that kept users engaged.

Apparently, this system tended to show people a lot of clickbait: thumbnails proliferated and misleading titles. Users would click, but then feel trapped, perhaps a little irritated, and then abandon videos partially afterwards. Ultimately, YouTube realized its user experience was disappearing and changed strategies.

2012: Watch time (a.k.a. view duration)

In 2012, YouTube announced an update to the discovery system tailored to identify the videos people really want to watch. This saw YouTube focus more on videos that hold attention throughout and further allowed YouTube to assure advertisers that it was offering a valuable, superior experience for people.

At the same time, YouTube was encouraging YouTubers to stop tricking the algorithm optimization — For example, creating longer videos to gain more watch time or making them shorter to get higher retention rates.

As it is today, YouTube encouraged them to create videos that people really want to watch.

Machine learning (a.k.a. the algorithm)

In 2016, YouTube released a whitepaper that shocked many. In it, their product engineers described the role of machine learning and deep neural networks in the platform’s recommendation system.

Though the whitepaper wasn’t a tell-all, what we do know is that YouTube tracks viewers’ apparent satisfaction to create a personalized and addictive stream of recommendations.

YouTube neural networks and machine learning infographic
(Source: Deep Neural Networks for YouTube Recommendations, 2016)

2016-2020: Borderline content, demonetization and brand safety

Since 2016 to date, YouTube has received a lot of questions regarding the types of videos their algorithm promotes and those it doesn’t. According to YouTube CEO Susan Wojcicki, YouTube takes its responsibilities seriously and tries to balance a broad, fair range of opinions while still while doing its best to make sure dangerous information isn’t spread. For example, a 2019 update to its algorithm resulted in 70% less watch time for “borderline” content. Borderline content is content that doesn’t violate the platform’s community guidelines but is misleading or harmful.

This is a serious issue because it touches every issue. For example, in March this year, content creators alleged that YouTube was demonetizing videos that so much alluded to the existence of the coronavirus.

Fast forward, irrespective of where you stand, it is worth noting the developments are still ongoing and thus both advertisers and content creators must keep tabs on what is happening. For example, if you are a creator, you need to aware that YouTube can turn around and demonetize your channel or videos if content crosses the line. 

Similarly, advertisers need to realize that the algorithm in its current form is designed to demonetize borderline content, mostly to protect brands. Besides, YouTube says it might never be able to guarantee 100% brand safety.

How does the YouTube algorithm work in 2020?

YouTube says their algorithm is a “real-time feedback loop that tailors videos to each viewer’s different interests.” It has the capability to decide what videos it suggests to every user.

Basically, it is designed to find the right video for every viewer and get viewers to keep watching. That means it is ever watching users’ behaviors while at the same time watching video performance.

The two main places the algorithm impacts are recommendation streams and search results.

How the algorithm influences search results

If keen you may have noticed there is a variation between the videos you get and what your friend gets when you search the same thing. For example, when you search “fastest cars in the world.” This is because search results are based on factors such as:

  • Video’s engagement (comments, likes, watch time)
  • Video’s metadata (description, keywords, title) and how well those match the user’s query

Does YouTube algorithm influence the recommended videos?

Yes, it does. First, it ranks them by assigning them a score based on performance analytics data. It also matches videos to people based on their watch history, and what similar people have watched.

Basically, the goal here is to match viewers with videos that they want to watch and to ensure they spend as much time as possible on the platform and see as many ads as possible.

YouTube also uses different metrics and signals to rank and recommend videos on the different sections of its platform. That said, let’s take a look at how the algorithm decides to show content to users on their suggested videos, search, home, subscriptions section, and trending.


Your videos’ search rankings are affected by two factors: relevance and keywords. When ranking videos in search, YouTube considers how well your content, titles, and descriptions match each users’ queries. The number of videos a user has watched from your channel, and the last time he or she watched other videos surrounding the same topic as your video are also considered. 

Home and suggested videos

YouTube always strives to serve the most personalized and relevant recommendations to each of its viewers. To achieve this, YouTube analyzes users’ activity history and find videos that could be relevant to them.

Next, they rank these videos by how well each video has engaged and satisfied similar users. They also consider how often each viewer watches videos from each channel or other videos surrounding the same topic. Finally, YouTube takes into account how many times YouTube has already shown each video to users.


The trending page is normally is a feed of popular and new videos in a user’s specific country. YouTube wants to balance innovation with popularity when they rank videos in this section. As such, they largely consider view count and rate of view growth for each video they rank.


YouTube has a subscription page where users can view all the recently uploaded videos from the channels they subscribe to. But this page isn’t the only benefit channels get when they acquire a ton of subscribers.

How does YouTube determine the algorithm?

Well, we don’t work at Google but based on the various factors in various public discussions over the years, it is evident that when it ranks a video, the algorithm looks at the performance: 

  • How much time people spend watching a video (retention or watch time)
  • How much time people spend on the platform after watching a video (session time)
  • Whether people click on a video (impressions vs. views: thumbnail and title are important, here)
  • How many likes, comments, dislikes, or shares a video gets (engagement)
  • How new a video is (note new videos may get extra attention in order to give them a chance to snowball)
  • How quickly a video’s popularity snowballs or doesn’t (this is called view velocity, rate of growth)
  • How often a channel uploads new videos

When it matches a video to a potential viewer, the algorithm looks at personalization:

  • Which topics and channels have they watched in the past?
  • How much time do they spend watching?
  • What don’t they watch?
  • How many times has this video already been surfaced for this person?
  • What have they engaged with in the past?

Under the Reach Viewers tab, you can view these metrics, which together demonstrate YouTube’s new prominence on watch time and click-through rate:

  • Impressions: How many times your video thumbnails were shown to viewers as a recommended video, in search results, or on the homepage.
  • Traffic sources for impressions: Where on YouTube your video thumbnails were shown to potential viewers.
  • Impressions click-through rate (CTR): How often users watched a video after seeing your thumbnails (based on logged-in impressions).
  • Watch time from impressions: Watch time that originated from people who saw your videos and clicked them on YouTube.
  • Views from impressions: This measures how often viewers watched your videos after seeing them on YouTube.

Rethinking “clickbait”: The relationship between watch time and click-through rate

You have probably heard about YouTube’s war against clickbait as the platform was flooded with confusing video thumbnails and exaggerated titles trying to beat the algorithm. 

The result: the attention shifted towards watch time as the key signal for ensuring the quality of a video. Many content creators reacted to this by abandoning the tactics that once helped them grab attention while competing against the hundreds of hours of video content uploaded every minute.

Unfortunately, that didn’t work either.

In a Q&A about prioritizing signals in the YouTube algorithm, one of the same Google engineers from the paper we mentioned earlier admitted, “It’s constantly a struggle because mostly you’re combating abuse at the same time. So if you optimize for click-through rate, you get clickbait, and if you optimize for watch time, you get incredibly long videos.”

If a video has a high click-through rate but generates low watch time, then it is clickbait without a doubt. But if attention-grabbing titles and thumbnails get people to click through and watch your videos, then that is ideal.

In a nutshell, if you want to get more views through YouTube’s recommendation engine, you need to optimize your videos and channel for both click-through rate and watch time.


YouTube algorithm will continue to change, and as it has been, Youtubers and brands will continue to be left wondering what to do next to stay on top of the game. However, despite the algorithm changing, YouTube’s core goal remains the same: to get more people watching and engaging with more videos on YouTube. Surprisingly, that is also your aim. Is that so!