There is no single “the algorithm.” YouTube runs separate recommendation systems for Home (Browse), Suggested videos, Search, and Shorts. They share one goal: predict what a specific viewer will watch and enjoy next. Every ranking decision is a bet on that prediction.
That means the algorithm doesn't rank your video. It ranks your video for one viewer at a time, using signals from everyone who watched before them.
The three signals that decide reach
Strip away the mythology and almost everything reduces to three measurable inputs:
- Click-through rate (CTR). When your thumbnail is shown, what share of people click? This gates everything: a video nobody clicks never gets the chance to prove itself.
- Watch time and retention. Once they click, how long do they stay? Average view duration tells YouTube whether the click was satisfied or baited.
- Engagement. Likes, comments, shares, and subscriptions after watching. These are the clearest “this was worth it” signals a viewer can send, and they weigh heavily in whether the system keeps testing your video on new audiences.
Velocity: why the first hours matter most
YouTube tests every upload on a small slice of your likely audience first. What happens in that test window decides the next push. A video that collects likes and comments quickly relative to its impressions reads as a hit, so the system widens the test. One that sits flat gets quietly shelved.
This is engagement velocity: not how much engagement you have, but how fast it arrives relative to views. Two videos with 1,000 views and 80 likes are not equal if one collected them in three hours and the other in three weeks. It is also why seasoned creators concentrate everything (community posts, premieres, replies) into the first hours after upload, and why some give the early signals a deliberate push.
What Browse and Suggested actually reward
Browse (the homepage) is a cold-start engine: it shows your video to people who haven't asked for it, based on how similar viewers responded. Suggested is the sidebar and end-screen row, driven by watch-session data: which videos keep people watching when queued after yours or your competitors'.
Both surfaces care about the same thing: satisfaction per impression. High CTR with strong retention and visible engagement is the pattern that earns impressions there. Search is the exception: it leans more on relevance (titles, descriptions, captions), which is why keyword-stuffed titles still rank in Search but die in Browse.
What doesn't work anymore
- Tags. Effectively decorative since 2021.
- Raw view spikes. Views without engagement or retention behind them don't move Browse. The system reads them as unsatisfied impressions.
- Upload frequency for its own sake. Consistency helps your audience form habits; volume alone doesn't earn recommendations.
The practical takeaway
Make the packaging (title + thumbnail) earn the click, make the first 30 seconds keep it, and then concentrate every engagement signal you can into the first hours after upload. The algorithm isn't a mystery. It's a feedback loop, and the loop is fastest at the start.
Related: do likes actually help rankings? and how to get more comments.