The Instagram Algorithm Explained: How Ranking Actually Works
What Instagram has publicly said about ranking on Feed, Stories, Explore, and Reels — the signals that matter, connected vs recommended reach, and what to do with it.
14 min read·By SocialTools Online·Published September 11, 2026
There is no single Instagram algorithm
Instagram has said publicly and repeatedly that the app does not run one algorithm. Feed, Stories, Explore, Reels, and Search each use their own ranking system, trained on different behaviour, because people use each surface for a different reason. You open Stories to check on people you know; you open Explore to find something new. Ranking the same way in both would make both worse.
This matters because most advice treats “the algorithm” as one thing you can satisfy with one tactic. In practice, a post that performs well in the feeds of people who already follow you succeeded under a different system than a Reel that reached strangers. Diagnosing performance starts with asking which surface delivered the views, and your own insights will tell you that directly.
The second consequence is that changes to one surface do not necessarily affect another. A quarter where your Reels reach collapses while your Stories completion stays healthy is not a shadowban; it is two independent systems producing two independent results.
The four categories of signal
Across surfaces, Instagram has described ranking inputs falling into broadly four groups. Information about the post itself: how many people engaged with it, when it was posted, its format and length, and any location attached. Information about the person who posted it: how interesting the account has been to people generally over recent weeks.
Then your activity: what you have liked, saved, shared, and watched, which tells the system what topics you want. And your history of interaction with a particular account: whether you comment on their posts, whether you have messaged them, whether you watch their Stories to the end. That last category is why a friend’s ordinary photo can outrank a polished post from a brand you follow passively.
From these inputs, each surface predicts a small number of actions — how likely you are to spend time on the post, comment, like, share, or tap through to the profile — and ranks accordingly. Which predictions get weighted most heavily differs by surface, which is the mechanical reason the same content performs differently in Feed and Explore.
Connected reach versus recommended reach
The most useful distinction for planning is between reach among people who already follow you and reach among people who do not. Instagram calls the second recommendations, and it has published separate guidance on what makes an account eligible for it. Treating these as one number hides the story your insights are telling you.
Connected reach depends heavily on your relationship with existing followers: whether they interact, whether they have muted you, how recently they opened the app. An account with a large but inactive following will see low connected reach no matter how good the post is, because the ranking system has learned those people do not engage.
Recommended reach depends more on the post’s early performance among a test audience of non-followers and on whether your account is eligible to be recommended at all. Instagram has said that recommendations exclude certain categories of content and that reposted or aggregated material — including content with visible third-party watermarks — is deprioritized in favour of original posts. If your reach among non-followers is near zero across many posts, originality and eligibility are the first things to examine.
What each surface tends to reward
Feed leans on your relationship with the account and how likely you are to spend time on the post. Carousels and longer captions do well here relative to elsewhere, because the surface tolerates and measures dwell time. Recency still matters, but chronology is one input rather than the ordering principle.
Stories lean almost entirely on relationship. Completion rate, taps forward and back, replies, and sticker interactions determine whether your Stories keep appearing near the front of the tray. Stories are rarely a discovery surface, which makes them the wrong place to expect growth and the right place to deepen the audience you have.
Reels lean on watch time and completion first, then on shares and sends to individual people. Instagram has said sends to friends are a particularly strong signal, which is intuitive: forwarding a video to someone is the most expensive endorsement available in the app. Explore and Search lean on topical relevance and on whether the content resembles material the viewer has previously engaged with.
Saves, shares, and sends over likes
The interactions that predict distribution best are the ones that cost the viewer something. A like takes a fraction of a second. A save is a decision to come back. A share to a Story is a public endorsement. A send in DMs is a personal recommendation to a specific person, which requires thinking about who would want it.
Practically, this reorders what content you should be making. Reference material — checklists, size charts, step-by-step explanations, comparisons — earns saves. Content that articulates something the viewer wants a specific friend to see earns sends. Content that makes the viewer look thoughtful or funny when shared earns Story shares. Content designed only to be pleasant earns likes and stops there.
It also reorders what you measure. If you judge posts by likes, you will systematically make more of your least distributable content. Pull saves, shares, and sends out of your insights and rank your last twenty posts by those instead; the ordering usually looks different from the like count, and the difference is instructive.
What the algorithm is not doing
It is not punishing you for posting too often or too rarely, in the sense of a global penalty. Posting frequency affects your results through the ordinary mechanism of having more or fewer posts to rank, and through audience fatigue if quality drops. There is no hidden counter tracking your cadence.
It is not applying a secret shadowban to ordinary accounts as a routine matter. Instagram does restrict recommendations for content that breaks its recommendation guidelines, and hashtags can be restricted, and accounts can be actioned for violations — all of which are visible in the app’s account status tools. Persistent low reach in the absence of any of that is much more often an audience or content problem than a punishment.
It is not counting your hashtags as a reach multiplier. Instagram has publicly downplayed hashtags as a distribution mechanism and pointed instead to keywords and content relevance. Hashtags still contribute to classification; thirty of them do not contribute thirty times as much.
How to actually use this
Separate your insights into follower and non-follower reach and track them independently. If connected reach is weak, work on the relationship: reply to comments and DMs, post Stories your audience actually responds to, and stop posting things your existing audience has demonstrated they ignore. If non-follower reach is weak, work on originality, topical clarity, and the first seconds of your Reels.
Pick one predicted action to optimize per post and design for it. A carousel built to be saved needs a first slide that promises reference value and a final slide worth returning to. A Reel built to be sent needs a clearly identifiable “this is so you” quality. Trying to maximize everything at once produces content that triggers nothing.
Then compare across ten to twenty posts rather than reacting to single results. Ranking systems are probabilistic and individual posts are noisy. The signal is in the pattern, and the pattern is only visible if you keep a record.
Where to verify this yourself
Instagram publishes its own explanations of ranking through its blog and through the head of Instagram’s posts and videos, and it maintains published recommendation guidelines describing what content is not eligible to be recommended. Those are the primary sources, they are free, and they are more reliable than any third-party reverse-engineering.
Read them once a year, because the surfaces change. A tactic that worked when Reels launched may be irrelevant now, and much of the persistent folklore about Instagram — engagement pods, posting-time formulas, hashtag counts — originates from advice that was already outdated when it went viral.
Then check your own account status and insights rather than trusting a diagnosis from a reach-checking site. The app tells you whether your content is eligible for recommendations and whether anything has been actioned. That is the authoritative answer to most “am I shadowbanned” questions.
Tools for this guide
Put these tips into practice with free utilities — no signup required.
Browse all tools →Related guides
About the author
SocialTools Online
Site editorial
Guides and tool instructions are written by the SocialTools Online team. We document how each browser tool works, platform limits we can verify, and practical workflows for creators and small businesses. We do not sell growth services or guarantee search rankings.
Cite or share this guide
Link the canonical URL above. Embeddable counters and citation HTML: Press & embed.