Last updated Feb 23, 2022
Instagram Feed is the primary location to access content from friends, family, businesses and people you’ve connected with. "When we first launched in 2010, Instagram was a single stream of photos in chronological order. But as more people joined and more was shared, it became impossible for most people to see everything, let alone all the posts they cared about. … So we developed and introduced a Feed that ranked posts based on what you care about most."
Feed ranking prioritizes multiple pieces of content to help people see the posts they’re most likely to find interesting, or are most likely to interact with. The Feed ranking system predicts how likely you are to comment on it, like and save it, or tap on a profile photo. “The more likely you are to take an action, and the more heavily we weigh that action, the higher up a post will be ranked in Feed.” Check out the Instagram blog where we shed more light on this.
First, the system gathers potential posts—excluding advertisements—from accounts you follow, like posts from friends or creators. It then removes the posts that violate our community guidelines .
Using what’s left, the system predicts how likely you are to interact with a post. To do this, it collects attributes from the post along with additional information like how often you interact with the author of a post. Based on those attributes, the model predicts how likely you are to like, save, tap or perform an action like watching a video. High model outputs (i.e., probability) indicates a higher likelihood that you’re interested in the post.
The system combines these likelihoods into a single numerical score for each post. It consistently weighs and combines these model outputs in the same way for everyone.
At this stage, the system repeats steps 1-3 for all other post types, including:
Shopping posts
Videos
Reels
Hashtags you follow
The system then normalizes the scores in each post type so that one type doesn’t dominate another.
Now that they’ve all been scored, the system demotes posts that contain misinformation identified by third-party fact-checkers, or posts shared from accounts that have repeatedly shared misinformation in the past.
Until now, each post has been scored individually. In this step, the system applies additional rules to ensure your feed contains a wide variety of posts, and one type of content does not dominate. For instance, we’ve created a rule to show no more than three posts in a row from the same account. These rules are tested to make sure that they positively impact our users by providing diverse content that aligns with their interests.
Lastly, the system combines all of the posts together and delivers them to your feed.
Using the information below, try ranking a hypothetical user's feed to see how it compares with what the feed system might predict.
Step 1: Select a profile
First, you’ll need to understand their preferences. Review the following three profiles and think about how you’d rank the posts in their feeds.
When you’re ready, select the profile you’d like to work with.
We understand people are confused about how our systems work. In an effort to provide more transparency, we’re introducing tools that help people better understand what they’re seeing and why.
In the Instagram Feed, we are testing the ‘Why you’re seeing this’ feature on posts from the accounts you follow. By tapping the overflow menu (...) on a post and selecting this option, you'll get detailed information about why we showed you a specific post. For example, you might see that Instagram is showing you a post from a friend higher in your feed because you like their posts more often than others.
How you use Instagram heavily influences the things you see and don’t see. You help improve the experience simply by interacting with the profiles and posts you enjoy, but there are a few more explicit things you can do to influence what you see:
You can select your close friends for Stories. This was designed as a way to let you share with just the people closest to you, but we will also prioritize these friends in both Feed and Stories.
You can mute an account if you’d like to stop seeing what they share, but are hesitant about unfollowing them entirely.
To help us demote posts you want to see less of, you can ‘Hide’ them. When you choose this option, the current post will be hidden, and posts similar to those will show up lower in your feed.
In addition to feed recommendations, we’ve been experimenting with ‘Favorites.’ This is a way for you to decide whose posts you want to see sooner. We’re also working on an option called ‘Following,’ which allows you to see posts from people you follow in chronological order.
If you see an ad you don’t like, there are a few things you can do to remove the ad from your feed, and to help influence the ads you see.
This diagram shows a list of the systems, ML models and other elements that make up the Instagram Feed ranking system.
Action
The activities—such as ‘like’ or ‘tap’—performed on an ad, page, app or event.
Attributes
Metadata about a post which serves as inputs into an AI Model, such as the time a post was created or how many comments it has.
Close Friends
A person or group of people that you choose to share content with privately.
Demote
The action of moving a post lower or to the bottom of an Instagram feed.
Feed
A continuously updated list of advertisements and recent posts from the people you follow on Instagram.
Interests
A measure of what a user cares about, based on past engagement and activity.
Model
A statistical representation of tasks learnt from datasets, which can provide meaningful insights and be used for predictions.
Normalize
Adjusting values measured on different scales to a common scale.
Probabilities
The likelihood that something will or won’t happen.
Rule
A business decision logic.
Score
The predicted numerical output of an AI Model.
System
A group of ML models, AI and non-AI technologies that work together to accomplish specific tasks, such as ranking posts in a feed.
Transparency
The act of sharing information about an AI system, model or decision.
Foundational models
Latest news
Foundational models