My Name Is Kalam

How the YouTube Algorithm Works in 2026

Last reviewed . Every claim links to its official source. By Muhammed Abdul Kalam

YouTube's recommendation system learns from 80 billion signals daily. For the homepage, it primarily uses your watch history. For Up Next, it uses the video you are currently watching. Videos are ranked by viewer personalisation and content performance, not by subscriber count or publish time.

YouTube does not promote videos to audiences — it finds audiences for videos. The system ranks each video based on two categories: how well it is personalised to a specific viewer, and how well it performs when offered to viewers. Every claim on this page comes from official YouTube Help or the YouTube Creator Blog.

Key signals

Official

YouTube's recommendation system learns every day from over 80 billion pieces of information it calls signals.

YouTube, 2021
Official

For Home recommendations, YouTube's system primarily relies on your watch history to decide which videos to surface.

YouTube, 2024
Official

YouTube uses classifiers to identify borderline content and demotes it in recommendations rather than removing it entirely.

YouTube, 2021

Key takeaways

  • YouTube finds audiences for videos — it does not promote videos to audiences. Performance per viewer matters more than total subscriber count.
  • Home recommendations use watch history; Up Next uses the currently-playing video as the primary signal.
  • The system learns from over 80 billion signals daily, including clicks, watch time, survey ratings, shares, likes, and dislikes.
  • Borderline content is demoted in recommendations, not removed — a 70% watchtime drop was reported after 2019 changes.
  • Publish time does not affect long-term video performance in recommendations.
On this page
  1. Ranking signals: viewer personalisation
  2. Ranking signals: content performance
  3. What gets suppressed or demoted
  4. Official creator guidance
  5. Where these signals come from
  6. FAQs

Ranking signals: viewer personalisation

For Home recommendations, YouTube's system primarily relies on your watch history to decide which videos to surface.

Source: YouTube, How YouTube recommendations work (2024)

Official Ranking signal Older data

For Up Next recommendations, YouTube uses the video you are currently watching as the main signal when suggesting the next video.

Source: YouTube, How YouTube recommendations work (2024)

Official Ranking signal Older data

Ranking signals: content performance

YouTube uses satisfaction surveys — asking viewers to rate videos — to measure satisfaction beyond just watch time as an input to recommendations.

Source: YouTube, How YouTube recommendations work (2024)

Official Ranking signal Older data

In 2011 YouTube discovered that clicking on a video does not mean you actually watched it, and shifted to prioritising watch duration over click-through rate.

Source: YouTube, On YouTube's recommendation system (2011)

Official Algorithm change Historical

What gets suppressed or demoted

YouTube uses classifiers to identify borderline content and demotes it in recommendations rather than removing it entirely.

Source: YouTube, On YouTube's recommendation system (2021)

Official suppression Older data

After YouTube began demoting borderline content in 2019, watchtime from non-subscribed recommended borderline content dropped 70% in the US.

Source: YouTube, On YouTube's recommendation system (2019)

Official Algorithm change Older data

Official creator guidance

YouTube officially advises creators to prioritise consistent quality content over a high frequency of uploads.

Source: YouTube, YouTube's Recommendation System (2024)

Official Platform recommendation Older data

Where these signals come from

YouTube

YouTube algorithm: FAQs

Does subscriber count affect YouTube recommendations?

No. YouTube's recommendation system does not use subscriber count as a direct ranking signal. Videos are ranked based on viewer personalisation and per-video performance data, not channel size.

What is YouTube's most important ranking signal?

Watch history combined with how well a video performs when offered to viewers. The YouTube blog confirms the system learns from over 80 billion signals daily, with watch history and satisfaction surveys being primary inputs.

Does publishing at a certain time boost YouTube recommendations?

No. YouTube's official guidance states that publish time is not known to impact a video's long-term performance in recommendations.

How does YouTube handle borderline content in recommendations?

YouTube uses classifiers to identify borderline content and demotes it in recommendations rather than removing it entirely. This approach resulted in a 70% drop in watchtime on recommended borderline content in the US in 2019.

Does taking a break from posting hurt YouTube recommendations?

No. YouTube officially states that the algorithm does not penalise creators for taking time off, and that a sustained presence is advisable but a single break will not hinder a channel's potential.