stellar
YouTube Algorithm in 2026: How It Works and What Changed
The YouTube algorithm ranks videos by predicted viewer satisfaction: clicks, watch time, surveys and feedback. Here is how it works in 2026 and what changed.
Key takeaways
- The YouTube algorithm ranks videos by predicted viewer satisfaction, not by upload time or subscriber count.
- Home, Suggested, Search and the Shorts feed each rank videos separately with different signals.
- Shorts live or die on the first seconds: the viewed vs swiped away share and rewatches matter most.
- Most 2026 changes are product and metric changes, such as longer Shorts and the new Shorts view count, not a new ranking goal.
- YouTube Studio's reach, retention and Shorts reports show where a video is losing out before you change anything.
- Bought engagement does not trigger recommendations, and YouTube has a fake engagement policy against it.
The YouTube algorithm is a set of recommendation systems that choose videos for each viewer. It uses what that person watched, clicked, skipped, liked and reported in surveys. It ranks videos by predicted viewer satisfaction, not by upload time or subscriber count.
In 2026 this applies to both long videos and Shorts. The difference is that Shorts are judged mostly on whether people stop swiping and keep watching. The sections below explain each surface, the signals that matter, recent changes, and the reports in YouTube Studio that show how your videos are doing.
How does the YouTube algorithm work?
YouTube does not push a video to everyone at once. It shows the video to small groups of viewers it thinks might like it, then watches what they do. If those viewers click, keep watching and come back, the system shows the video to more people who look similar.
YouTube puts it as "the algorithm follows the audience." The system is not judging your video against an ideal. It is predicting, viewer by viewer, whether this person will be glad they watched. The official How YouTube Works page on recommendations lists clicks, watch time, survey responses, sharing, likes and dislikes as inputs.
Each surface runs its own ranking with its own goal. A video can do well in Search and get almost nothing from Home, or the other way round.
- Home: a personal feed based on watch history, subscriptions and what similar viewers enjoyed. For most established channels this is the largest source of traffic.
- Suggested videos: the column next to a video and the autoplay queue. It favours videos often watched in the same session as the current one.
- Search: relevance to the query comes first, so titles, descriptions and the spoken content matter more here than anywhere else.
- Shorts feed: a vertical, swipe-based feed where the main question is whether viewers stop and watch.
- Subscriptions and notifications: chronological for subscribers who check that tab. Notifications only reach subscribers who turned on the bell, and YouTube may still limit them.
Which signals does the YouTube algorithm use?
The main signals are click-through, watch time and retention, viewer satisfaction, and direct feedback. No single one decides ranking. A high click rate with poor retention usually stalls, because YouTube reads it as a title that over-promised.
YouTube has explained this in public several times. A post on the official YouTube blog about its recommendation system describes moving from raw clicks to watch time, then adding satisfaction surveys and "not interested" feedback. Raw views stopped being the target long ago.
| Signal | What YouTube is measuring | What you control |
|---|---|---|
| Impressions click-through rate | How often a shown thumbnail gets clicked | Thumbnail, title, topic choice |
| Average view duration and retention | How long viewers stay and where they leave | Hook, pacing, editing, video length |
| Satisfaction surveys | Whether viewers rate the video as worth their time | Delivering what the title promised |
| Likes, dislikes, comments, shares | Direct reactions after watching | Clear prompts, content that people want to discuss |
| "Not interested" and "Don't recommend channel" | Negative feedback that suppresses future recommendations | Avoiding misleading packaging |
| Session behaviour | Whether viewers keep watching YouTube afterwards | End screens, playlists, series content |
| Viewer history and context | Topics, device, time of day, language | Consistent topics so YouTube can find your audience |
Do subscribers, upload frequency or video length matter?
Subscribers matter less than most creators think. A subscriber is just a viewer with a strong past signal, and if they stop watching, YouTube stops showing them your uploads. Upload frequency has no direct effect, though more uploads give the system more chances to find a hit.
Length is neutral. A 6-minute video with strong retention can beat a 25-minute video that loses half its viewers in the first minute. Pick the length the topic needs, then remove anything that does not hold attention.
How does the YouTube Shorts algorithm work in 2026?
The Shorts algorithm decides whether a viewer stops scrolling and watches, then whether they finish, replay, like, share or subscribe. Shorts are shown in a feed where the viewer did not choose your video, so the first second carries far more weight than it does in long form. Shorts can now run up to three minutes, but length only helps if people stay.
Shorts also draw from a different pool than long videos. A channel with a large long-form audience does not automatically get Shorts reach, and the reverse is also true. Watching a Short can lead some viewers to a channel's long videos, but YouTube treats the two formats as separate recommendation problems.
What Shorts signals matter most?
- Viewed vs swiped away: the share of people who stayed instead of swiping on. YouTube Studio shows this for each Short.
- Average percentage viewed: Shorts can go above 100% when viewers loop the video.
- Engaged actions: likes, comments, shares, remixes and subscribes from the Shorts player.
- Topic consistency: Shorts on similar topics help the system learn which viewers to test you with next.
How did Shorts view counting change?
In 2025 YouTube changed how Shorts views are counted. A view now counts each time a Short starts or replays, with no minimum watch time. The older, stricter number became a separate metric called engaged views, which you can still see in YouTube Studio.
This changed what the public view count means, not how ranking works. Recommendations still depend on retention and satisfaction, not the headline number. For monetization thresholds and how engaged views fit in, see our guide to YouTube Shorts monetization requirements.
What changed in YouTube algorithm updates for 2026?
YouTube rarely announces "algorithm updates" the way creators expect. Most changes arrive as product features, new metrics or policy updates, and the ranking goal stays the same: predicted satisfaction. The useful question is which recent changes affect how your videos get tested and measured.
Changes from the past two years that still shape results in 2026:
- Longer Shorts: Shorts can now run up to three minutes, so retention and pacing matter more than fitting under a minute.
- Shorts view counting: the switch to counting plays and replays, with engaged views kept as a separate metric.
- Thumbnail testing: Test & Compare in YouTube Studio lets eligible channels test up to three thumbnails. YouTube picks the winner by watch time share, not click rate alone, which tells you what it values.
- Inspiration and research tools: YouTube Studio's Inspiration tab and the Research tab in Analytics show topics your viewers search for and what content gaps exist.
- Stricter handling of repetitive and mass-produced content: YouTube Partner Program rules on reused and repetitive content affect monetization. Low-effort, template content also tends to get weak satisfaction signals.
Where can you follow YouTube algorithm news?
Use first-party sources and treat everything else as opinion. The most reliable places are the Creator Insider channel, the official YouTube blog, the "What's new" panel in YouTube Studio, and the YouTube Help Center changelog. Many claims about "the 2026 algorithm change" online come from one creator's traffic drop, not from anything YouTube said.
If your views drop suddenly, check three things before blaming an update. Look at whether impressions fell (fewer tests), whether click rate fell (packaging), or whether retention fell (content). Each points to a different fix.
