How do video apps choose what to recommend?

Kohaku
KohakuWhen you open a video app, it is full of videos you want to watch. How are they chosen?
The answer in 30 seconds
KohakuWatch time, survey answers, and people who watch like you are the clues.

Video apps line up recommendations by predicting which videos you are likely to be happy watching, based on what you have watched before, how long you watched, likes and dislikes, and answers to satisfaction surveys. What people with similar viewing habits often watch is another important clue.

Kohaku
KohakuNow, let’s look at it in more detail.
A hand pointing a remote control at a menu on a TV screen
Choosing what to watch from a list on the TV screen with a remote control.Photo: Shixart1985 / CC BY 2.0 / Wikimedia Commons

Predicting videos you'll be glad you watched

According to YouTube, recommendations appear mainly in 2 places: the homepage you see when you first open the app, and the "Up Next" panel that shows while you watch. Using clues such as what you watch, likes, dislikes, subscriptions and survey answers, it predicts videos you will want to watch and feel were worth it. Netflix also says it estimates how likely you are to enjoy a title from your viewing history and ratings, members with similar tastes, and information about the title such as its genre, actors and release year.

Similar viewers point to your next video

One big clue is "people who watch like you." In a 2021 explanation, YouTube said that if people who watched the same tennis videos as you also enjoy jazz videos, you may be recommended jazz videos, even if you have never watched a single one. Back around 2008, when it started, it was one big popularity page, ranked by view count and the same for everyone.

Figure: Recommendations from similar viewers
Watched videos, watch time, likes and dislikes, star surveys and similar viewers all feed one box that re-ranks the recommendations
  1. 1Videos you have watched
  2. 2Clues such as watch time, likes and dislikes, and surveys
  3. 3The system that gathers the clues and re-ranks videos
  4. 4Videos watched by other people with similar habits
  5. 5The re-ranked recommendations

From clicks to watch time and surveys

The way of measuring what people really watched has changed, too. In 2011, YouTube learned that clicking on a video doesn't mean you actually watched it, so from 2012 it also began using watch time. Right after that change, views dropped by 20%, but it chose to put viewers' value first. Today it also uses surveys that ask you to rate a video from 1 to 5 stars, and only time spent on videos rated 4 or 5 counts as "valued watch time." For people who don't answer, a model trained on the answers that were collected fills in the gap.

Figure: Signals for measuring satisfaction
Clicksa click doesn’t meanyou actually watchedWatch timeused since 2012how long you watchedSurveysrate from 1 to 5 stars4 or 5 stars count
Kohaku
KohakuA click alone can't tell whether you really watched. Even the measuring takes some thought, doesn't it?

You can pause or delete your history

You can also shape your recommendations yourself. On YouTube, you can pause, edit or delete your search and watch history at any time. And for topics where accuracy matters, such as news and health, it also uses ways to show videos from trustworthy sources first.

Want to know more?

  • According to YouTube, more viewing comes from recommendations than from channel subscriptions or search (as explained in 2021).
  • On Netflix, titles you watched recently count more than ones you watched long ago. Netflix also says it does not use age or gender in its recommendations.
  • On the Netflix homepage, the most strongly recommended titles start on the left of each row. If your language is set to Arabic or Hebrew, they start on the right.
Kohaku
KohakuRecommendations are predictions of what you'll enjoy. Tidy your history, and the lineup may change too.

Quick quiz: test yourself

KohakuRemember what you just read? Give it a try!

What is a big clue for recommendations?

Show the answer
  1. What similar viewers watch
  2. Today's weather
  3. Only the viewer's age

What similar viewers watch Videos often watched by people with similar viewing habits may be recommended, even if they are a kind you have never watched before.

Sources
  1. On YouTube’s recommendation systemYouTube Official Blog
  2. YouTube のアルゴリズムに基づくおすすめ機能YouTube(How YouTube Works) (in Japanese)
  3. How Netflix’s Recommendations System WorksNetflix Help Center

This article was written after checking the sources above as of 2026-10-03. Explanations may change as research progresses. Photo credits are shown below each photo.

Next fun fact ▶

TechnologyHow does a smart speaker know when you call it?Call its name, and a smart speaker responds right away. How does it notice?
NANI?! Animal Encyclopedia
Our book is on AmazonNANI?! Animal Encyclopedia

40 surprising animal facts with big photos and short, easy text. A fun-facts book for kids and grown-ups to read together.

See it on Amazon