The YouTube algorithm is critical to creators’ success, but this doesn’t make it any less difficult to understand. As we all know, YouTube and Google don’t divulge the secrets of their algorithms. They do, however, give you guidelines to follow which can tip the algorithms in your favor.
Even if YouTube won’t reveal its algorithm’s nuts and bolts, you can still discover them through old-fashioned research. Here, we’ve done all of the heavy lifting for you: Read on to learn everything you need to know about the YouTube algorithm for more views, subscribers and audience engagement.
Table of Contents
- Why Does the YouTube Algorithm Matter?
- The YouTube Algorithm’s Illustrious Past
- The Videos YouTube Loves
- Flash in the Pan or Long-Term Success?
- Tips for Lasting YouTube Relevance
- Five YouTube FAQs Answered
Why Does the YouTube Algorithm Matter?
As one of the premier social networks for videos, any SEO would be remiss not to include YouTube marketing in their digital marketing stack. Because you’re on iloveseo.com, we think you either have some interest in pursuing YouTube marketing, and you may be slightly familiar with how a typical search engine works:
- A user submits a query.
- The algorithm returns a selection of results it believes are relevant to that query.
- The user chooses a result to click on.
While modern search engines have added machine learning, human quality raters and user feedback into the mix, that basic formula remains.
But on YouTube, the algorithm works differently. While it may behave similarly to a standard search engine, it also adds content curation to the mix through its auto-generated recommendations.
It does so by displaying videos it thinks users will be interested in on both the homepage and in the platform’s sidebar:
The YouTube algorithm doesn’t just answer users’ needs—it proactively addresses them before they type so much as a single keyword. Of course, this assumes that you are subscribed to the platform and logged in.
That’s why YouTube SEO is unique: You’re aiming to get your videos in front of users who may not have made a query to begin with.
The YouTube Algorithm’s Illustrious Past
As of January 2021, YouTube is the second most used social media platform in the world, surpassed only by Facebook. The site also gets billions of views per day—by comparison, it was receiving 100 million daily views in July 2006.
But YouTube was no small fry, even upon its first release in 2005. Started by three PayPal employees, the site quickly saw its user base skyrocket and was purchased by Google just 18 months later.
Despite Google’s involvement, YouTube’s recommendation system remained rudimentary for years. The site didn’t begin recommending related videos until 2008, and even then it did so based on simple metrics such as click count and watch time.
The tide turned in 2016 with the implementation of machine learning. As explained in a white paper from Google, YouTube’s neural networks provide recommendations by analyzing each user’s history and context alongside other, video-specific metrics:
Although most of us may not understand exactly how it works, you only have to watch an hour or two of YouTube videos to see the results: An almost-endless, highly-personalized supply of videos presented to you at every turn.
The formula clearly works—at the time of this writing, more than 500 hours of content are uploaded every minute. The site boasts over two billion logged-in users a month, and YouTube is used in more than 100 countries.
Translation? YouTube presents one of the best opportunities to connect with your audience, and you’re seriously missing out if you don’t take advantage of it.
The Videos YouTube Loves
If you want to get on the YouTube algorithm’s good side, start by analyzing the types of videos it already favors.
YouTube’s own documentation states that its algorithm “follows the audience”. It does so by observing users’ viewing habits, including:
- what they watch and don’t watch;
- their watch time;
- likes and dislikes; and
- any “not interested” feedback previously given.
When users are actively watching a video, the algorithm will also recommend videos that are topically related, uploaded by the same channel or similar to other videos the user has already watched.
While the algorithm isn’t perfect (remember when it was recommending that everyone watch videos about 1,000 degree knives?), it is effective at identifying content users will likely find engaging.
All that is to say, if users engage with a video—either by viewing it, subscribing to its parent channel, liking or even disliking it—then the algorithm is more likely to recommend it to other viewers.
Flash in the Pan or Long-Term Success?
As with the aforementioned 1,000 degree knife videos, having one video heavily recommended by YouTube’s algorithm doesn’t necessarily mean you’ll enjoy lasting success.
After all, if they’re going to keep giving you views and clicks, users have to be just as interested in the rest of your channel’s videos as they were in the one originally recommended to them.
Take for example Jennelle Eliana, a van-dwelling YouTuber who amassed over one million subscribers after uploading just two (yes, two) videos. One year later, those two videos are still the most popular on her account, with 18 million views on one and ten million on the other:
But most importantly to Eliana’s long-term success, each of her videos still gets hundreds of thousands of views—at the time of this writing, most new uploads garner one million views or more. Her subscriber count has grown as well, sitting at nearly 2.5 million as of April 2021.
The lesson here is: by consistently creating unique and engaging content, YouTube channels can turn 15 minutes of fame into a steady stream of interested viewers.
Remember, quality is always better than quantity when it comes to long-term success. While you probably shouldn’t release a video every few months, don’t feel pressured to release one every day either.
Tips for Lasting YouTube Relevance
If you are looking to take advantage of the many benefits of YouTube’s algorithm, it’s a good idea to pay attention to the following guidelines:
In the eyes of the algorithm, a video’s engagement is a better indicator of its quality than its raw view count. Engagement includes views, but it also takes into account likes, dislikes, comments, share, watch time and the like. Remember to encourage your viewers to like, subscribe and share your video non-intrusively.
First, navigate to the traffic sources section in YouTube Analytics. There, you’ll see a list of the words people are using to find your channel. Be sure to incorporate those words into your videos’ titles and descriptions. While you’re at it, brush up on your keyword research skills to proactively identify important keywords.
Write Intriguing Titles and Descriptions
In the competitive world of YouTube, an eye-catching video title is as good as gold. Plus, your videos’ descriptions are a perfect place to optimize for keywords, send viewers to certain links and add even more valuable content. YouTube recommends that your titles be 60 characters or less and your descriptions be one to two paragraphs.
Create Custom Thumbnails
Stand-out thumbnails are an indispensable way to draw in more viewers. To make yours as effective as possible, add custom thumbnails that clearly show users what your video is about.
Add a Strong Call to Action
When you use calls to action (CTAs) at the end of your videos, you’re increasing your chances of getting more views, engagements and conversions. Luckily, YouTube makes it easy to do with cards and end screens.
Cross-Promote Your Videos
Don’t just rely on YouTube’s internal system to drive views—instead, put your social media marketing savvy to good use and link to each new video from other platforms.
Five YouTube FAQs Answered
Who better than YouTube’s own team members to answer your most burning algorithm-related questions? In an October 2020 video from Creator Insider, two employees did just that:
Don’t feel like watching the whole thing? Here’s a transcript for your convenience:
If one of my videos under-performs, is that going to hurt my channel? Could a few poor videos pull down better videos in the future?
I would say, don’t worry about our systems making some decision about your channel being good overall, we’re always going to be following the audience. And so if it doesn’t matter to the audience that your last video wasn’t as good as maybe some of your other videos, it’s not gonna matter to us either.
We want each video to get out to the audience that’s gonna watch it and enjoy it. And we often see that channels produce some great videos and some videos that aren’t as great, and so we’re very careful not to have all of our recommendations driven by like channel reputation. We look at each video’s performance. And so this is what enables some videos to go viral and other videos not to. When you see variations in your impressions or in your views, this is the system listening to the audience, and when the audience responds well we reflect that video level of performance.
So unless you’re really impacting how your audience is feeling about your channel, and whether the audience is going to stop watching your channel in the future, I wouldn’t worry about variability or having a few bad videos. Know that your next upload, if you produce something that is really engaging to the audience, we’re gonna wanna get it out there as much as possible.
Is there a point at which the number of videos per day/week on each channel is so high that the algorithm is overwhelmed and videos slip through?
No, there’s no limit on how many videos can be recommended to a given viewer from a channel in a single day. You can upload as much as you want, and how much viewers are willing to watch depends on their individual preferences. Our recommendation system will likely continue recommending them as long as viewers are willing to watch them.
But if you are uploading more videos than usual, and you start noticing that they’re receiving fewer views per video, that is an indicator you may have exhausted your audience and uploaded too much that they’re not willing to watch. So that is something to keep in mind as you experiment with how many videos you upload.
Batch uploads can actually be a really effective strategy at getting people to watch more. We’ve seen some creators release series like on a single day, which then gives viewers an opportunity to watch through them because they’re all available at once. So it really depends on your channel and your strategy, keep experimenting.
Now, although there’s no limit for recommendations, there is for notifications. So if you upload more than three videos in a given 24 hour period, the notifications limit is that viewers can get a maximum of three new video notifications per channel in that 24 hour period. And that was designed as not to overwhelm viewers who are more likely to turn off all notifications from YouTube, if they get overwhelmed with notifications. So yes, for notifications there is a limit, no for recommendations.
My channel has been around for quite a few years and I think I may have lots of inactive subscribers, should I create a NEW channel and then re-upload the videos in order to appear more acceptable to the algorithm?
So inactive subscribers, are they gonna hurt your performance? I think the answer is gonna be probably not, because if you remember how we decide about recommendations, we’re looking at each viewer when they come to YouTube, what do we know about that viewer and how can we rank the videos such that the videos that they’re most likely to watch and enjoy are showing up at the top of their homepage and other recommendations?
And so for a viewer that subscribed to your channel, let’s just say five years ago, and hasn’t watched in the past four years, our system is gonna know that this viewer hasn’t watched your channel in a while and your channel isn’t going, your new upload isn’t going to probably be the best candidate for that viewer.
And so you’re less likely to get a recommendation for those inactive viewers, which means that, in places like the homepage and other recommendations, the viewers who are going to see your upload are the ones that are active on your channel and interested in watching.
And so I wouldn’t worry about inactive subscribers hurting your performance because those inactive subscribers are less likely to see your latest uploads.
Now, you may say, “Oh, but what about the subscriptions feed?” Yes, if they go to the subscriptions feed they may still see your uploads. But we do most of our learning about content performance based on the surface that we’re doing the recommendations on. So if we’re trying to produce recommendations on the homepage, we’re mostly looking at how does that video perform when we show it on the homepage.
So that information from the subscriptions feed about, you know, a viewer saw an impression and didn’t click on, isn’t going to have a strong influence at all on the homepage recommendations which are learning primarily from the homepage. So unless you wanna like really rebrand your channel and just start with something completely separate, I wouldn’t really start a [new] channel because your old channel probably still has a base of subscribers who would be interested in your new uploads compared to zero on a new channel.
And so I would start with the base that I have, and not worry so much about whether inactive viewers are going to hurt your performance.
How important is external traffic for driving recommendations?
So external [traffic] is interesting. Some channels receive more or less traffic from external sources, and it is a signal that we consider which is really helpful in identifying videos that became really popular outside YouTube. So if your video starts getting more external traffic, it can certainly help kick off the process of getting your video recommended.
That being said, once your video is there in recommendations showing up in suggested or on the homepage, our systems are gonna learn from how viewers engage with it when it is recommended on those surfaces. So do viewers continue to click, watch it through and enjoy those videos when they’re being recommended? So if yes, viewers find it interesting when it’s recommended, it will continue to be ranked even more highly on that particular surface.
So external is something we consider, but the long-term success of your video is gonna be how it performs when it’s recommended to viewers.
I’m getting lots of traffic from external websites which is causing my click-through-rates and average view durations to drop, is this going to hurt my video’s performance?
Sometimes you’re gonna get traffic from sources where the viewers are gonna be less familiar with your content and your channel. So naturally, they’re gonna be less likely to click or watch [for] quite as long.
So if you get more traffic from ads or external [sources], sometimes you’ll notice your videos’ overall click-through rate or average view duration will kind of tank a little bit. It’ll be much lower than usual.
That’s not a problem in that in discovery, each individual surface like home and suggested have their distinct ranking models. So on home, home is learning from when a video is recommended on home, do viewers choose to watch it? How long do they watch and do they enjoy it when it is recommended there?
So overall click-through rate and average view duration sometimes are not as indicative of videos’ performance, and it won’t hurt your videos’ long-term success because the recommendation system, again, is learning how is this video doing when it is recommended to viewers.
So don’t worry about lower click-through rates and average view durations if you’re getting more views from a traffic source like an external, it shouldn’t hurt your videos’ success.
The YouTube Algorithm Is Your Friend
When you’re struggling to get your videos to perform how you want them to, it’s easy to get frustrated with YouTube’s algorithm and its seemingly mysterious ways.
But, when you know how the algorithm works and why it recommends some videos over others, your path is made clear. More viewers, higher CTRs (click-through rates) and higher conversion rates are all within your grasp—all you need to do is put in the legwork to achieve them.
Screenshots by author / April 2021
Google / September 2016