How YouTube Recommendations Work and Change
What YouTube publicly says about how recommendations work, the long-term direction of its changes and how to diagnose a drop in views.
"The algorithm changed" is one of the most common explanations for a drop in views, and one of the least useful. YouTube adjusts its systems constantly, but the direction has been steady for years: it tries to show each viewer the videos they are most likely to watch and be satisfied with. This article explains what YouTube publicly says about how recommendations work, how to tell a real change from normal variation, and how to adapt without guessing.
What YouTube says it measures
On How YouTube Works: recommendations, YouTube lists the kinds of signals its system uses:
- Clicks: whether a viewer chooses a video when it is shown.
- Watch time: how long viewers watch, rather than whether they clicked.
- Survey responses: viewers are asked to rate videos, which helps measure satisfaction beyond watch time.
- Sharing, likes, dislikes and "not interested": direct feedback on whether a viewer valued a video.
- Viewing history and context: what each viewer has watched before and what they are watching now.
Search uses a different balance, centerd on relevance to the query, engagement and quality, as described on How YouTube Works: search.
The long-term direction
Across the changes YouTube has described over the years, the pattern is consistent:
- A move away from rewarding clicks alone, which encouraged clickbait.
- Then away from raw watch time alone, towards measures of whether viewers were satisfied.
- More weight on personalization, so the same video can do very differently for different audiences.
- Separate feeds and habits for long videos, Shorts and live streams.
The practical meaning is that tactics designed to trick the system tend to stop working, while videos that serve a clear audience keep getting distributed.
Is it the algorithm or the video?
Before blaming a system change, check the usual causes. YouTube's performance FAQ and troubleshooting page covers many of them:
- Seasonality: many topics rise and fall with holidays, school terms or events.
- A strong previous video: a hit raises the baseline, so the next upload can look weak in comparison.
- Topic or format change: viewers who came for one kind of video may not want another.
- Packaging: a weaker title or thumbnail lowers click-through rate and therefore distribution.
- Audience fatigue: repeating the same idea too often.
- Policy issues: check for notices about account standing or limited ads.
How to diagnose a drop
- In YouTube Analytics, compare the last 28 days with the previous period and with the same period last year.
- Split views by traffic source. A drop in browse features points to recommendations; a drop in search points to the topic or the competition.
- Check impressions first. Fewer impressions means less distribution; similar impressions with fewer views means lower click-through rate.
- Look at retention on recent videos using key moments for audience retention.
- Note what changed on your side: schedule, format, length, presenter or topic.
How to adapt
- Serve a defined audience. A clear promise makes it easier for the system to find the right viewers.
- Package honestly. Titles and thumbnails that set the right expectation lead to satisfied viewers.
- Deliver early. Get to the point in the first 30 seconds.
- Build sessions. Use playlists, end screens and series so viewers keep watching.
- Treat Shorts separately. Measure them on their own terms and link them to related long videos.
- Change one thing at a time so you learn what made the difference.
Where to follow official changes
YouTube announces product and policy changes on its official channels, including the YouTube Creators site and help pages. Treat claims of secret algorithm updates from unofficial sources with caution unless they are backed by data.
Common questions
Do views drop after breaks?
Often, because subscribers and recommendations have moved on to other channels. Views usually recover as you publish consistently again.
Are smaller channels penalized?
No. Recommendations follow how viewers respond to each video, so a small channel's video can be widely recommended.
Does deleting videos help?
Rarely. Improving weak videos or leaving them alone is usually better than deleting them.
Should I copy trending videos?
Learn from their packaging and structure, but keep to what your audience came for.
Next steps
For the search side of the picture, read our YouTube SEO guide. If you want help diagnosing a drop, see our YouTube SEO services.
Sources
- YouTube: How YouTube Works: recommendations
- YouTube: How YouTube Works: search
- YouTube Help: YouTube performance FAQ and troubleshooting
- YouTube Help: Get started with YouTube Analytics
- YouTube: YouTube Creators
Was this article helpful?
Written by
