A definition in plain words
When people say the algorithm, they mean the software that picks and orders what appears on your screen. Meta's own description is more specific: content is selected, ranked and delivered by an AI system, in which several machine learning models work together, and those models and their inputs change frequently as the system learns.
Two consequences follow. There is no fixed formula to learn, because the models keep changing. And the system is not one thing: Meta publishes a separate explanation, which it calls a system card, for each part of Instagram.
One app, several systems
| Part of the app | What it ranks | Where candidates come from |
|---|---|---|
| Feed | Posts from accounts you follow. | Accounts you follow. Suggested posts and ads are handled by separate systems. |
| Stories | The order of the stories tray. | Accounts you follow, without ads. |
| Explore | The grid of recommended photos and videos. | Authors you engaged with, and media similar to media you engaged with. |
| Reels | Which reels come next, and in what order. | Accounts you follow, and reels similar to ones you recently engaged with. |
So a post can do well in the feeds of your followers and never appear in Explore, or the other way around. They are different systems asking different questions.
From thousands of posts to one screen
Each system works as a funnel. It gathers candidates, removes what breaks the Community Guidelines, narrows the rest with a light model, and then scores the survivors with heavier ones. Meta gives approximate sizes for three of them:
The final order comes from a score. Posts the system predicts will be most valuable to you go first, and the feed also tries to keep a balanced mix of content types.
What the algorithm predicts about you
Every system card lists the predictions it makes and the signals behind them. A few from each:
- Feed: whether you will swipe through a whole carousel, enjoy a post (like, comment, reshare, time spent), skip it, visit the author's profile, or send it to someone in a direct message.
- Stories: whether you will tap a story at the top of your feed, reply to it, and whether you and the author are close, judged partly by the messages you exchange.
- Explore: whether you will follow the author, look at a post for more than five seconds, watch more than 95% of a video, comment, or tap Not Interested.
- Reels: whether you will watch less than three seconds, use the audio in a reel of your own, tap Interested, or comment.
The DM prediction is worth noticing for anyone who posts. Its signals include whether the post is informative, such as a tutorial, a how to or a product review, and whether it was created within the last 90 days.
What you can change about what you see
The cards list the controls each system listens to. In Feed: mute, add to favorites, save, hide a post, unfollow, report, and the Following view, which shows accounts you follow in time order without recommendations. In Explore: Not Interested. In Reels: Following, for reels in time order from accounts you follow.
Hiding a post also limits similar content, and reporting tells the system to show you less of the same kind. If most of what you are shown is off, Instagram also offers a full reset, covered in how to reset your Instagram algorithm. How the feed itself is built, favorites included, is in what Instagram feed means.
What the system cards do not say
Much of what circulates as algorithm advice is not in any of these documents. The cards name predictions and signals; they do not give weights, a best time to post, a magic number of hashtags or a penalty for editing a caption. Treat any rule that is not on Meta's own pages as somebody's guess, however confident it sounds.
The cards also say the opposite of a fixed formula: the models and their signals change frequently. A tactic that worked once is evidence about that moment, not a law.
What it means for your own posts
Read the predictions from the other side and they describe a good post. People stay on it, swipe to the end, send it to a friend, and do not skip it in the first seconds. None of that can be faked for long, but all of it can be designed for:
- Make the first seconds count, because skipping within two seconds and watching under three are named signals for Reels; nine ways to open a reel is about exactly that.
- Give people a reason to send it, such as a useful how to or a review, which matches the signals behind the DM prediction.
- Earn the full swipe on a carousel, since swiping through the whole thing is a prediction of its own.
Then check what happened in Instagram analytics, post by post.
For a brand account
For a brand, the useful takeaway is that there are several audiences inside one app. Followers meet you in Feed and Stories, ranked mostly on their history with you. Strangers meet you in Explore and Reels, ranked on how people like them responded to similar posts. A content plan that only serves one of these is leaving the other on the table.
If you want a read on which of these is working for your account, the free social media audit looks at it post by post.
How we will test this
Not run yetEverything above comes from the published sources listed below, and nothing here is presented as something we measured. Our own test on live accounts has not been run yet.
This is the test we will run: we follow a fresh test account's Feed, Explore and Reels for two weeks while engaging with one topic, use Not Interested and Hide on another, and record how each part of the app changes and how fast.
Platforms change behavior without announcing it, so once the first test is run we repeat it every quarter, and this section will say when it was run and on which devices and app versions.
Sources
Read on October 1, 2026.
- Meta Transparency Center, updated June 29, 2026, Instagram Feed AI system
- Meta Transparency Center, updated September 18, 2025, Instagram Stories AI system
- Meta Transparency Center, updated June 22, 2026, Instagram Explore AI system
- Meta Transparency Center, updated November 11, 2025, Instagram Reels Chaining AI system
Questions people ask next
What does the Instagram algorithm mean?+
The AI systems that choose and order what you see on Instagram. Feed, Stories, Explore and Reels each have their own.
Is there one Instagram algorithm?+
No. Meta publishes separate system cards for Feed, Stories, Explore, Reels and other parts of the app, each with its own predictions.
How does the Instagram algorithm decide what I see?+
It gathers candidate posts, removes ones that break the Community Guidelines, predicts how likely you are to engage with each, and ranks them by score.
How many posts does the algorithm consider?+
Meta gives rough numbers: Explore retrieves up to 1,500 items and narrows them to 100, Feed selects about 700 posts and scores about 500, Reels selects and scores about 100.
What signals does the Instagram algorithm use?+
Many, including who posted, the type of content, your history with the author, time spent, likes, comments, shares, saves and skips. Each system card lists its own.
Does the algorithm change?+
Yes. Meta says the models and their input signals change frequently as the system learns.
Can I see posts without the algorithm?+
Yes. The Following view shows posts from accounts you follow in time order, without recommendations. Reels has a Following option too.
What does Not Interested do?+
It tells Explore or Reels you do not want more of that kind of content, and is one of the signals the system predicts.
Is there a best time to post for the algorithm?+
Not in Meta's system cards. They describe predictions about engagement, not a posting hour.
Do shares matter to the Instagram algorithm?+
Sharing a post in a direct message is one of the Feed predictions, with signals including whether the post is informative and recent.
Can I reset the Instagram algorithm?+
Yes. Reset suggested content under Content preferences clears your suggestions, though it does not change who you follow.