Facebook Feed AI system

UPDATED JUN 4, 2025
The content you see on your Facebook Feed is selected, ranked and delivered to you by an artificial intelligence (AI) system. Within one AI system, multiple machine learning models work together to deliver your experience. These models and their input signals are dynamic and they change frequently as the system learns and improves over time.

Overview of Facebook Feed

When you view and interact with Facebook, one of the underlying AI systems delivers the connected content you see in your Feed. This is content you’ve chosen to see, such as status updates, photos, videos and other content from people, Pages and groups you follow or are connected to on Facebook.
Note: Your Feed might also include suggested content—such as new Pages to follow and groups to join—as well as advertisements. That content is not powered by the AI we describe in this system card. Suggested content is described in Facebook Feed Recommendations.
How Facebook Feed works
The AI system behind Facebook Feed automatically orders the posts in your Feed by predicting what you will find most valuable and relevant. Here’s how it works:
  1. Gather inventory
    First, the system gathers all potential posts shared by friends, Pages you follow and groups you've joined, which excludes posts flagged for going against our Community Standards.
  2. Leverage signals
    Next, the AI system considers a variety of input signals about each post. These signals might include who created the post and how you previously interacted with them, whether the post is a photo, a video or a link, or how many of your friends liked the post. A lightweight model is run to select approximately 500 of the most relevant posts. The system also applies certain integrity processes to help reduce the distribution of problematic content.
  3. Make predictions
    From there, the AI system has models that help it make predictions about content you’ll find most relevant and valuable.
  4. Rank posts by score
    Finally, the system calculates a relevance score for about 500 posts and puts them in order by this score. Posts that the system predicts will provide more value for you are shown higher in your Feed. The system also tries to ensure that your Feed has a balanced mix of content types. That means, for example, you wouldn’t see multiple video posts in a row, but rather a variety of post types.
How to customize what you see
Your experience on Facebook Feed is personalized based on your activity, and you have options to control or customize what you see. Below, we describe how to do this with different in-product features. Options shown here may not be available to everyone.
Hide
You can hide a post so you won't see that post again. This action also helps to minimize similar content from showing up in your Feed.
Reconnect
You can reconnect to follow a person, Page or group that you unfollowed in the past.
Snooze
You can temporarily stop seeing posts from a person, Page or group. You can also restart, stop or add more time to your snoozes.
Report content
If you see something you think goes against our Community Standards, you can report it to us. You can also report posts that you think are spam or false news.
Interested/Not interested
This feature lets you customize what you see in your Feed. Clicking "Interested” or “Not interested” on a on a post will temporarily increase or decrease the ranking score for a post and other posts like it.
Unfollow
To stop seeing certain posts, you can unfollow a person, Page or group.
Manage favorites
You can select people and Pages you want to prioritize. This means their posts will be shown higher in your Feed, and you'll see their newest posts first.
See newest content first
To view content from people, Pages and groups you follow in reverse chronological order, visit the Feeds tab. Learn more.
How the AI delivers content to you
How likely you are to scroll past a post rather than engage with it
Signals influencing this prediction include:
  • At which position is the post located on your News Feed?
  • Percentage of posts posted by your friends compared to total posts viewed.
  • Number of times you have seen posts with the same content type (e.g. video, photo, etc) as the current post in the past hour
  • Average time spent on each post
  • Number of times you quickly scroll past the same friend’s post in the past 7 days
How likely you are to scroll past a post rather than engage with it, taking into account additional information about the user's preferences when interacting with different post formats
Signals influencing this prediction include:
  • Number of seconds user spent on status update posts in the past 14 days
  • Total time user spent viewing Feed
  • The number of comments in the post
  • Number of times user watched a video for more than 30 seconds in the past 14 days
  • Number of times user viewed status update posts in the past 14 days
Likelihood that a user will skip a Story In Feed (SIF)
Signals influencing this prediction include:
  • Number of stories that the user has seen in their Feed over the past 28 days.
  • Number of videos that were viewed during a 2-hour window
  • Number of videos that were skipped during a 2-hour window
  • Number of stories in Feed that user has interacted with over past 28 days
  • A weighted count of the number of times a post has been viewed in 2 hours.
Likelihood a user will skip a video channel or not
Signals influencing this prediction include:
  • Number of times a user has watched a video for a certain amount of time in the last 7 days.
  • How far down in the user's Feed the first link story appears, compared to the first story they saw.
  • Number of times a user has engaged with sports-related content in the past 30 days
  • Number of times a user has clicked on a video to enter the video channel in the last 7 days.
  • How far down the user’s Feed the first video story appears compared to the first story they already saw.
How likely you are to click 'Interested' if the interested/not interested options are presented to you below the post
Signals influencing this prediction include:
  • List of owners the user has clicked the interested control in last 180 days
  • List of posts the user has used "Interested" control in the past 180 days
  • The current time
  • How many times user watched a video for 30 seconds or more in the last 30 days
  • The length of the video
How likely you are to click into a story in Feed and enter full screen view
Signals influencing this prediction include:
  • How many stories you have watched in the past 28 days
  • How many stories you have watched in the past 7 days
  • How many unique stories you have watched in the past 7 days
  • How many original stories you have watched in the past 28 days
  • How many unique stories not posted by you have you watched in the past
Whether a user will click into and view the list of users that have liked and/or reacted to that post
Signals influencing this prediction include:
  • Number of times a specific topic or category has been associated with a post or content that helps identify its topic.
  • Number of interactions the post has received.
  • Number of times a user interacts with a post or content, where the user is the main person mentioned or tagged.
  • Average length of comments on a post.
  • A measure of how relevant the comments on a post are to the topic of the post.
Predicts whether a user has a deep/strong intent to engage with posts from a group they are connected to
Signals influencing this prediction include:
  • total user's actions (e.g., like, comment, etc) on group users interact with divided by 3
  • Number of times user saw a particular type of post (e.g. photo, video, link).
  • Percentage of comments user made on posts from a specific group, compared to the total number of comments made on all posts in recent history.
  • Amount of time that has passed since a user joined a particular group.
  • Historical data of user's actions (e.g., like, comment, etc) on group users interact with in the past one day
How likely the user will click on a link that takes the user to a website outside of Facebook and predicts how long the user will stay on that website
Signals influencing this prediction include:
  • Percentage of times user clicked on a link that takes to a website outside of Facebook, compared to the total number of times user saw such links in your Feed over the past 14 days
  • The total number of times user saw a post in their Feed over the past 14 days, specifically that are shared or saved by others.
  • Percentage of times user clicked on a link that takes to a website outside of Facebook, compared to the total number of times user saw such links in your Feed over the past 14 days, specifically for posts that are shared or saved by others.
  • How often user will interact with a post on their mobile device.
  • The total number of times user saw a post in their Feed over the past 14 days.
Likelihood of a user having a strong intent to interact with their friends/group on Facebook
Signals influencing this prediction include:
  • Quickly predicts how likely you are to interact with a post or activity from someone you follow, based on how much your friends have engaged with similar content.
  • User interactions with stories in the past.
  • How likely you are to interact with a post or activity from someone you follow, based on how much your friends have engaged with similar content.
  • Interactions between the user and author of post.
  • Interactions between the user and author of post in a public setting