How the YouTube Algorithm Works
There is no single “YouTube algorithm.” There are several recommendation surfaces — Home, Suggested, Shorts, and Search — each scoring videos a little differently. What unites them is the goal: predict which video will satisfy a particular viewer, then recommend it. YouTube has long optimised for watch time and increasingly for genuine satisfaction, not raw views.
The surfaces, briefly
- Home. Recommends from across the platform based on a viewer’s history and what similar viewers watched and finished.
- Suggested. The sidebar/next-up — recommends videos related to what someone is watching now and likely to keep them watching.
- Search. Behaves like a search engine: relevance to the query, plus engagement and satisfaction signals.
- Shorts. A separate short-form feed optimised for quick, swipeable consumption and completion.
The signals that matter
Across surfaces, the heavy signals are watch time and retention (do people keep watching, and how far), click-through rate (do titles and thumbnails earn the click), and satisfaction (engagement plus, in YouTube’s case, actual viewer surveys). The combination matters: a strong thumbnail that earns clicks but loses viewers quickly is a worse signal than a plain one that holds them. That is why clickbait underperforms over time — it wins the click and loses the retention.
What it means for how you make videos
YouTube rewards videos that keep a promise. A clear title and thumbnail set an expectation; the video has to deliver it quickly and hold attention. Because watch time compounds, long-form rewards depth and structure, while Shorts rewards a single sharp idea. This is also why repurposing works well on YouTube — one strong long-form video becomes Shorts, a community post, and a search asset, eachsuited to its own surface.
The honest version
The exact models change and YouTube does not publish weightings. But the durable truth holds: YouTube recommends what it predicts will satisfy a specific viewer, judged mostly by whether they watch, stay, and come away satisfied. Make videos worth finishing, set honest expectations, and let the surfaces do their job.
Frequently asked questions
Does YouTube rank videos by views or by satisfaction?
By predicted satisfaction more than raw views. YouTube uses watch time, retention, and signals of whether viewers were genuinely satisfied (including survey feedback) to decide what to recommend. A video with fewer views but strong retention can out-recommend a higher-view, low-retention one.
How is the YouTube Shorts algorithm different from long-form?
Shorts is a separate recommendation surface optimised for quick, swipeable consumption, so it leans on completion and rewatch like other short-form feeds. Long-form leans more on watch time, retention, and click-through from titles and thumbnails. They are different engines on the same platform.
Do thumbnails and titles affect the YouTube algorithm?
Indirectly but strongly. Titles and thumbnails drive click-through rate, and the algorithm pairs that with whether people stay once they click. A high click-through with poor retention (clickbait) underperforms a clear title that delivers on the promise.