YouTubePublished Updated

How the YouTube algorithm actually works

YouTube recommends videos to each viewer from clicks, watch time, likes, shares and satisfaction surveys. What YouTube has said about each surface.

The short answer

YouTube recommends videos one viewer at a time. Its own explanation lists the signals: clicks, watch time, survey responses, sharing, likes and dislikes. Watch time is weighted as “valued watch time”, judged by asking viewers in surveys whether a video was worth their time, because a click alone does not mean someone watched, and watching does not mean they were glad they did.

Recommendations are most of YouTube. Hootsuite’s 2026 guide puts them behind about 70% of what people watch, through the Home feed, the Suggested videos beside and after a video, and Shorts. A video that earns clicks and then holds and satisfies the people who click is shown to more people like them.

What each part of YouTube uses

Each surface weighs the signals differently. This is how the 2026 guides from Hootsuite and Sprout Social describe them.

What each part of YouTube uses to choose videos, from Hootsuite’s and Sprout Social’s 2026 guides
SurfaceWhat it uses
HomeThe video’s performance (click-through rate, watch time, engagement) and the viewer’s watch and search history, time of day and device
Suggested videosThe topic of the video being watched and the viewer’s history, to keep the session going
SearchHow well the title, description and video match the search, and how well it fits the viewer’s interests, plus engagement
ShortsHow often viewers watch rather than swipe away, how much they watch, replays, likes, comments and shares; click-through rate does not apply

Clicks open the door; satisfaction keeps it open

YouTube’s 2021 explanation tells the history. It first counted clicks, then added watch time in 2012 because a click did not mean a viewer watched, and later added surveys because watching did not mean the viewer valued it. Sprout Social’s 2026 guide notes that YouTube collects feedback through millions of these surveys, with questions such as “Was this video helpful?”, and that a video people finish but do not find satisfying may be shown less.

For a creator, this means the title and thumbnail have to make a promise, and the video has to keep it. A thumbnail that over-promises can win the click and lose the recommendation.

Shorts are judged differently

Shorts appear in a feed people swipe through, so there is no click to measure. Sprout Social names the share of viewers who watch rather than swipe away as the top Shorts signal, and Hootsuite adds how much of a Short people watch, replays and loops, and the usual likes, comments and shares. The first second of a Short does the job a thumbnail does for a long video.

What that means for a video

Turn the signals into a checklist:

  • An honest promise. A title and thumbnail that describe what the viewer will get earn clicks that turn into satisfied watching.
  • A strong start. Start with the thing the title promised, so the people who clicked for it stay.
  • A reason to keep watching. Watch time counts, so pace the video so each part leads to the next.
  • Words for search. Search matches the title, description and the video itself to what people type, so use the words they would use.
  • A next video. Suggested videos keep a session going, so a series gives YouTube an obvious next video to suggest.

What costs a post reach

The commonest way to lose reach on YouTube is to win the click and lose the viewer. YouTube added watch time and then satisfaction surveys precisely because clicks alone did not mean people valued a video, so a title or thumbnail that promises more than the video delivers can earn views at first and fewer recommendations after. Slow openings and padded runtimes do the same in a quieter way.

Questions

Do subscribers matter to the YouTube algorithm?

Less than many think. YouTube says its system can suggest a channel’s videos to someone who has never watched it, by comparing their habits with similar viewers. What decides a video’s reach is how the people it is shown to react, subscribers or not.

Does the time I upload matter?

YouTube’s Todd Beaupré has said time of day and device are among the signals the recommendation system uses, as quoted by Sprout Social. Publishing a little before your viewers are usually on YouTube, which YouTube Studio shows, helps a video’s first hours.

Is click-through rate or watch time more important?

Both, in order. Clicks get a video watched, and watch time and satisfaction decide whether it keeps being recommended. YouTube added watch time precisely because clicks alone rewarded videos people did not watch.

Does YouTube punish clickbait?

Its signals work against it. A click without watching counts for little, which is why YouTube added watch time in 2012, and its 2021 post says it began demoting sensationalistic tabloid content in 2015.

Do likes and dislikes matter on YouTube?

Yes. YouTube lists both among its signals, with clicks, watch time, survey responses and sharing.

Does the algorithm favour longer videos?

It favours watch time viewers value, not length. A long video people leave early, or watch without enjoying, does worse than a shorter one they finish and rate well.

Are Shorts and long videos recommended the same way?

No. Shorts have no click to measure, so they are judged on whether viewers watch rather than swipe, how much they watch, and replays, as well as likes, comments and shares.

Sources

  1. YouTube, On YouTube’s recommendation system, September 2021.
  2. Hootsuite, How the YouTube algorithm works, October 2026.
  3. Sprout Social, YouTube algorithm: how it works and tips to optimize, May 2026.

See where your viewers stop watching.

Connect YouTube and PostEase shows each video’s views and audience retention as YouTube reports them, next to your other channels. Free to start.

Start free