How the Instagram Algorithm Actually Works
There isn't one algorithm, there are several, and they optimise for different things in different places in the app. What Instagram has said publicly, what follows from it, and what to ignore.
Few subjects generate more confident, unsourced writing than Instagram ranking. Most of it describes ranking factors in precise numerical terms that nobody outside Meta could know, and much of it contradicts the next article.
What follows sticks to two things: what Instagram has said publicly through its own channels, and what reliably follows from how recommendation systems work. Where something is uncertain, it says so.
There is no "the algorithm"
The most useful correction first. Instagram has explained repeatedly through its creator resources and its company blog that different surfaces in the app use different ranking systems, because they are doing different jobs.
- Feed shows you mostly accounts you already follow, ordered by predicted interest.
- Stories shows accounts you follow, weighted heavily toward ones you interact with.
- Explore shows you almost entirely accounts you do not follow, chosen from content that performed well with people whose behaviour resembles yours.
- Reels is largely recommendation-driven and, like Explore, is not constrained to your following.
This distinction explains the most common confusion in the category. "My reach dropped" usually means one surface changed, not all of them. An account whose Feed reach is steady while Explore reach collapses has a very different problem from one losing followers.
It also locates where growth comes from: your existing audience sees you in Feed and Stories; new audiences find you through Explore and Reels. Content designed for one is often wrong for the other.
What the systems are predicting
Recommendation systems rank by predicting the probability that you will take some action, then ordering by those predictions. Instagram has described signals along these lines:
Information about the post. How recent it is, its format, its length, and early signals of how people are responding.
Information about the author. How often people who see this account's content engage with it. This is the closest thing to an "account quality" score, and it is relative rather than absolute.
Your history with that author. Whether you follow them, whether you have interacted before, whether you consistently watch their content to the end. This dominates in Feed and Stories, and is close to irrelevant in Explore.
Your activity generally. What you have engaged with recently, which builds the profile used to find people whose behaviour resembles yours.
Why sends and saves matter more than likes
A consistent theme in what Instagram has said publicly: not all engagement weighs the same, and the heavier signals tend to be the more effortful ones.
The logic is straightforward once stated. A like costs a fraction of a second and is often reflexive. Sending a post to a specific person requires deciding it is worth someone else's attention and choosing who. Saving requires deciding it is worth returning to. Both are costly, and costly actions predict genuine interest far better than cheap ones.
This has a practical consequence that survives most ranking changes: content built to be forwarded outperforms content built to be approved of. The question worth asking before publishing is not "will people like this" but "would anyone send this to a specific person, and who".
On "shadowbanning"
The term is used to explain almost any drop in reach and is usually the wrong explanation.
Instagram's published position is that content which does not violate its rules but sits close to them can be made ineligible for recommendation surfaces, while remaining fully visible to your own followers. That is a real mechanism, it is documented, and it is narrower than the folklore suggests: it affects Explore and Reels recommendations rather than hiding you from people who already follow you.
Most sudden reach drops have duller causes: a change in what you posted, a change in format mix, seasonal shifts in usage, or a single earlier post that performed unusually well and made the following weeks look worse by comparison. Before concluding you have been penalised, check whether Feed reach and Explore reach moved together. If Feed held and Explore fell, that is a recommendation-eligibility story. If both fell, it is more likely a content story.
Professional accounts can see account status and content eligibility in the app directly, which is a better source than any third-party checker.
What to ignore
Precise numeric ranking claims. "The first thirty minutes determine everything", "the algorithm favours a 7% engagement rate". Nobody outside Meta has these numbers, and they are stated with a confidence the evidence does not support.
Anything dated. Ranking systems change continuously. An article confidently describing this year's algorithm will be partly wrong within months, which is why it is worth going to Instagram's own material for mechanics and treating everything else as interpretation.
Engagement-bait tactics. Comment pods, follow-for-follow, engagement groups. These generate the cheap signals that carry least weight while producing an audience with no interest in what you sell.
What actually holds up
Five things that have survived every ranking change and will likely survive the next:
- Make things people forward. Sends and saves are the signals that matter and the behaviour that compounds.
- Match the format to the surface. Reels and Explore reach strangers; Feed and Stories serve the audience you have. Different jobs.
- Be consistent enough for the system to learn. Recommendation systems need repeated data on who responds to you. Sporadic posting gives them little to work with.
- Hold attention early. The opening seconds determine whether the strongest signal on video is ever generated.
- Post for a specific person, not a demographic. Content that is clearly for someone gets sent to that someone. Content for everyone gets sent to nobody.
Note that none of those depend on the current ranking mechanics being what anyone says they are. That is the point: build habits that are robust to the system changing, because it will.
Platform ranking systems change frequently and are not fully documented publicly. Mechanics described here reflect Instagram's own published material; verify current behaviour against their creator resources before making decisions that depend on it.