How the YouTube Algorithm Works in 2026
YouTube's recommendation system follows the audience, not the video. Here's how it decides what to recommend on Home, Suggested, Search and Shorts in 2026, and what creators should focus on.
Key takeaways
- YouTube recommends videos by predicting what each viewer will watch and be satisfied with.
- Key signals include click-through, watch time, retention, satisfaction surveys, likes, shares and 'Not interested' feedback.
- Home, Suggested, Search and Shorts each use the same principles with different emphasis.
- Creators win by making videos for a specific audience, with packaging that matches the content.
YouTube often says its algorithm “follows the audience.” It isn’t looking for the “best” videos in general; it is trying to predict which video each viewer will watch and be glad they watched. Understanding that changes how you make content.
Where recommendations appear
| Surface | How videos get there |
|---|---|
| Home | Personalised picks based on viewing history, subscriptions and similar viewers |
| Suggested (next to and after videos) | Videos related to what someone is watching or has watched |
| Search | Relevance to the query, plus engagement and quality signals |
| Shorts feed | A separate, swipe-based feed driven by watching and swiping behaviour |
The signals that matter
Click-through: do people choose to watch?
Impressions CTR shows whether your thumbnail and title make people want to watch. But a click alone isn’t success.
Watch time and retention: do they stay?
Watch time became YouTube’s core signal back in 2012, replacing raw views to stop clickbait. Today, how long people watch and how much of the video they watch (audience retention) both matter.
Satisfaction: were they glad they watched?
YouTube runs surveys asking viewers to rate videos, and uses likes, dislikes, shares, comments and “Not interested” feedback to estimate satisfaction. In 2026 YouTube added more direct mobile controls, “Not interested” and “Don’t recommend this channel,” which feed into this.
Context: what else does this viewer watch?
The same video can be recommended to very different people depending on their history. That’s why targeting a specific audience works better than trying to appeal to everyone.
Shorts work differently
On the Shorts feed, viewers swipe rather than choose from thumbnails, so the first second matters enormously. Signals like whether people watch to the end, rewatch or swipe away quickly carry a lot of weight. Shorts can be up to three minutes long since October 2024.
What has changed recently
- 2024: creators must label realistic AI-generated or altered content.
- July 2025: the “inauthentic content” monetisation policy replaced “repetitious content.”
- July 2026: clarified Partner Program guidance; Analytics renamed Insights; new viewer controls.
See the YouTube update history for the full timeline.
What creators should do
- Pick a clear audience and make videos for them.
- Package honestly: strong thumbnails and titles that match the content. See thumbnail and title tips.
- Earn the first 30 seconds: deliver on the promise quickly.
- Make viewers glad they watched. Satisfaction is the goal, not just watch time.
- Study your data in Studio Insights.
- Stay original. Mass-produced content is a monetisation and reach risk.
For a structured review, use the YouTube channel audit.