The YouTube algorithm in 2026: how videos actually get recommended
Home, Suggested, Search and Shorts each rank differently. Here is what each surface measures and which of those numbers you can move.
YouTube
People say "the algorithm" as if it were one system with a grudge. YouTube runs several rankers, and each one decides a different part of the app. Knowing which one you are trying to win changes what you should fix.
Home: personalised, click-driven
The Home feed predicts what a specific viewer will click and keep watching, based on their history. Your video gets a small trial audience of people who watched similar things; if the click-through rate and watch time on that trial beat the alternatives, Home widens the audience. This is why a video can sit at 200 views for two days and then take off: it finally won a trial.
Suggested: what plays next
Suggested videos sit beside and after the current one. The ranker looks for videos that viewers of the current video went on to watch. To appear there, make videos on the same topic as videos people already watch, and make them worth finishing, because Suggested traffic is judged on watch time above everything.
Search: relevance plus satisfaction
Search matches words in the title, description, transcript and chapters, then re-ranks by how satisfied earlier searchers were. Practical use: put the phrase people type into the title, say it out loud in the first minute so the transcript carries it, and add chapters with plain names.
Shorts: swipe or stay
Shorts are ranked almost purely on behaviour in the first second: did the viewer swipe away, watch to the end, rewatch, like or share. Subscriber count barely matters. A Short from a 50-subscriber channel can pass a million views if the swipe-away rate is low.
What you can move
- CTR: thumbnail and title. Test one change at a time; YouTube lets you A/B thumbnails now.
- Retention: cut the intro, deliver in the first 30 seconds, edit out every pause.
- Session: end screens to a strong related video; playlists that autoplay.
- Early velocity: views, likes and comments in the first 24 hours push the trial audience wider. Post when your audience is online and ask a real question in the description.
What does not work anymore
Tags are nearly irrelevant. Keyword-stuffed descriptions are ignored. Sub-for-sub rings fill your subscriber list with people who never watch, which drags down the numbers the rankers care about. The honest version of "gaming the algorithm" is giving a good video a strong first day, which is exactly what a modest, gradual push of views does, and nothing more.
