YouTube's algorithm has evolved over the past two decades and will continue to evolve as it adopts newer AI models that are more capable of predicting what viewers want to watch. YouTube isn't intersted in promoting all YouTubers. It will only promote content that enough viewers find worth watching. While this makes total sense, most YouTubers are at a complete loss when they invest a lot of time in producing a video and not see it gain any traction. In their eyes, the video bombed. This may not be the case. There are many very good videos on YouTube that have very low exposure. This will often happen because the channel itself has failed to reach certain thresholds that the algorithm requires to get more exposure.

As of 2024, there are nearly 4 million videos uploaded to YouTube every day. All those videos are competing for views. Yet there are limited resources in how far YouTube can recommend videos. First, there is the limited real estate on every user's Home feed. Then there are the actual servers that have to push video recommendations and search results. Only those videos that do exceptionally well and align with the interests of the viewer are going to show up near the top of search results.

In order to alleviate the problem of limited resources, YouTube places the responsibility of gaining traction on the YouTuber posting a video. Only when enough subscribers on a channel watch a new video posted by the channel will the algorithm deem it worthy enough for further promotion.

There are multiple levesl of promotion for videos. Once a video gets enough views from the channel subscribers, the algorithm will then promote it to a larger audience that it believes will be interested in it. These are people who have already watched similar videos. This is unlike the TikTok aglorithm that simply throws it out to a bunch of random people and then decides whether to promote it even more if it gets enough likes.

For better or worse, the YouTube algorithm takes a different approach. First, the algorithm notifies subscribers of a channel about a new video by placing a impressioon of it on their Home feed and providing notifications if the user enabled them. It then compares the total number of clicks to the total number of subscribers. It the percentage of clicks is high enough, the algorithm will perform additional analysis to deem if it's worthy for futher promotion. If it doesn't get enough clicks, it's game over. This process typically takes 24 hours due the fact that subscribers are located around the world and needs to give everyone a chance to have noticed that a new video has been posted.

Even having enough video views from subscribers isn't enough to guarantee further promotion. Those who watched the video had to have watched it long enough and it must have enough likes. If the video is uninteresting, or even worse, really bad or even clickbait, it will be watched for a very short time, and even get a lot of dislikes. A video that performs poorly with subscribers is not going to get promoted. For a video that is around 10 to 15 minutes long, most videos on YouTube tend to see viewers drop off around the 30 second mark. This doesn't mean it's bad. How many drop off by the time the 3 minute mark is reached will generally define whether a video is worth further promotion or not. The only way a YouTuber can know what the threshold is where the algorithm will decide to promote the video futher is to look at metrics of other videos they did that performed well.

Many YouTubers will attempt to change their thumbnail and title during the first few days if they see that their video is not getting the results that they expect. But this is a risky undertaking. The updated thumbnail and title may actually result in fewer views because maybe they aren't as engaging. YouTube does provide what is known as A/B testing of thumbnails, but as of November 2024, this feature is only available to a limited number of YouTubers who already have a large number of followers. With A/B testing, you define two thumbnails. You can't define two titles. YouTube then sends one of the thumbnails to one set of viewers and the other thumbnail to a different set of viewers. It then tracks to see which thumbnail resulted in more clicks over a duration of time and then uses only that thumbnail.

While A/B testing with thumbnails is something all YouTubers want, the title does play an important role as well - just not as important as the thumbnail. Users do look at the title when browsing videos but generally only when they see a thumbnail that they find interesting and need more context about the video from the title.

If the algorithm decides to promote a video beyond the subscribers, the video then enters a promotion phase that is made up of several levels. If the video does well at the lowest level, it is promoted to more viewers at the next level. If it does well on the next level, it gets promoted to even more viewers at the next level. Videos tend to only be promoted to any level for about a day. Regardless what level the video manages to achieve, YouTube will stop promoting the video after about 4 days. It doesn't stop abruptly. Instead, it begins to taper off on the number of impressions it issues. YouTubers see this happen all the time and misinterpret this to mean that their video has run its course and has gotten all the views it will ever get. This is certainly true for many, if not most videos. But even the best MrBeast video is going to decline within a week. With 4 million videos uploaded every day, YouTube's goal is to provide fresh content on a regular basis in order to keep viewers engaged.