Based on YouTube's own explanations, checked October 18, 2026: How YouTube recommendations work.
What does YouTube say it's optimizing for?
Viewer satisfaction. YouTube says it looks beyond clicks and total watch time, using surveys and feedback to estimate "valued" watch time, and models to predict satisfaction where it doesn't have survey data.
Where do recommendations show up?
| Surface | What drives it most |
|---|---|
| Home | The viewer's watch history and interests |
| Up Next / suggested | The video they're currently watching |
| Search | How well the video matches the search, plus satisfaction signals |
| Shorts feed | What similar viewers enjoyed |
What does this mean for a business channel?
- Serve one kind of viewer well. Clear focus helps YouTube find people like your best viewers.
- Keep the promise. A title that over-promises gets clicks and then dissatisfaction.
- Open strong and stay tight. You have about 30 seconds to show people they're in the right place.
- Don't chase everyone. A broad video that pleases nobody in particular rarely gets recommended.
What doesn't help?
Tag stuffing, misleading thumbnails and asking for likes every 30 seconds. None of these make viewers more satisfied.
Example
Worked example: why one video got recommended and another didn't
A financial planner publishes two videos in the same month:
- Video A: "Retirement Planning Explained." A broad title, a slow two-minute introduction, and a video that covers ten topics lightly. Many people click, most leave in the first minute, and few watch anything else afterward.
- Video B: "Can I Retire at 60 on My Savings? (Real Example)." A specific question, the answer previewed in the first 20 seconds, then a walkthrough. Fewer people click, but those who do watch most of it, leave thoughtful comments and click the end screen to a related video.
YouTube's description of its system suggests Video B is the better candidate for recommendations: the people who chose it appear satisfied, and there's a clear type of viewer to find more of. Video A's clicks don't help much if viewers aren't satisfied.
How to read your own signals in YouTube Studio
- Impressions click-through rate: are the right people choosing your video when they see it? A low rate suggests the title and thumbnail aren't clear or appealing to your audience.
- Audience retention: where do people leave? A steep drop in the first 30 seconds usually means the opening doesn't match the title's promise.
- Traffic sources: are views coming from Browse features (home), Suggested videos, or YouTube search? Growth in Browse and Suggested means recommendations are picking the video up.
- Returning viewers: a growing share means people come back for more, which is a strong sign of satisfaction.
Myths worth dropping
- "Posting time decides everything." Timing matters a little for the first hours; satisfaction matters for months.
- "Deleting and re-uploading resets the algorithm." You lose the video's history, comments and links.
- "Longer videos always rank better." A video should be as long as it needs to be to keep its promise, no longer.
- "Small channels can't get recommended." Recommendations are about each video and each viewer, not channel size alone.
A simple routine after each upload
- Check the first 30 seconds of retention after 48 hours.
- Read every comment for the first week; repeated questions are your next videos.
- Add the video to a playlist and link it from an end screen on a related older video.
Dave’s take
A principle I keep coming back to
"More important than the number of views is the number of engaged viewers." (Dave)
That's close to how YouTube describes its own goal.
AI shortcut
How can AI speed this up?
AI tools and plans checked October 18, 2026. They change often, so confirm on each tool's own page. No affiliate links.
- Diagnose a video. Ask Studio why a video's views changed, in plain language.
- Check your promise. Paste your title and first 30 seconds of transcript into ChatGPT and ask whether the opening delivers the title's promise.
Tools mentioned:
- Ask Studio in YouTube Studio: free, built into YouTube Studio on desktop for most creators.
- ChatGPT: free plan, Go ($8 a month in the US) and Plus ($20 a month).
AI is great for first drafts. Always edit the result so it sounds like you and is accurate for your business.
Key takeaways
- YouTube recommends videos it predicts each viewer will be satisfied with.
- Signals include watch history, surveys, likes, dislikes, shares and 'not interested'.
- Homepage suggestions lean on a viewer's history; Up Next leans on the current video.
- Make videos the right people genuinely value and keep your promise.
