Sledru

News explained

How Instagram ranks content, explained by its head

Adam Mosseri published the first video in a series about ranking. Three minutes, one unbroken take, no numbers at all, and one uncomfortable admission. Here is what he explained, plus the part the video left out.

Alexey Rusakov7 minutesupdated September 4, 2026

In short

Mosseri opened a video series on ranking, and the first instalment explains reasoning rather than tactics. There is no one algorithm: Feed, Stories, Reels and Explore each rank separately, with a split between follows and recommendations layered over them. A post’s slot comes from predicted likes, comments, sends and completions, and recency weighs enough to keep the feed nearly chronological. Viewers reshape their own feed through Favourites, interest topics and Not Interested. Falling reach is far more often ordering than punishment, and it is worth checking against facts rather than guesses.

What happened

On 2 September 2026 Adam Mosseri, the head of Instagram, posted a video running 2 minutes 49 seconds. One take, no cuts, filmed at a desk. He announced a series about why the platform ranks content and how it does it.

He opened with a warning: the comments would get spicy, and he knew it going in. Ranking has been a sore subject for creators since 2016, when the chronological feed gave way to sorting.

The value here is not new numbers. There are none. The value is that the mechanism gets explained out loud by the person accountable for it, with more videos promised.

The video hands out no growth recipe. No signal weights, no figures, no “do this and you will blow up”. Reasoning instead of instructions.

Why anything gets ranked at all

Mosseri named the reasons plainly, and most of them describe the platform rather than the viewer.

The first one is easy. Sorting helps people spend their time on what they actually care about instead of an undifferentiated stream.

The second is counterintuitive, and it concerns creators. An unranked feed fills up with whoever posts most often. Brands and businesses post more than creators. Creators post more than your friends. The app drifts away from friends and creative work towards news and publishers.

Then comes the argument he made twice. When posts are not ranked, people view fewer of them. Total views shrink, and the average reach of the average creator shrinks along with them. His phrase: the overall pie gets smaller.

And the admission worth staying for. Instagram’s own business gets smaller in that picture too. He said it himself, without waiting for the comments to say it for him.

  • An unranked feed rewards posting frequency, not the accounts people want to read.
  • Average reach drops for everyone because viewers get through fewer posts.
  • The app tilts away from friends and creators towards brands and publishers.
  • Platform revenue falls along with reach.

There is no single algorithm

Talking about “the algorithm” in the singular is the misconception Mosseri went after first. Instagram runs many ranking systems, built differently.

Feed, Stories, Reels and Explore each sort by their own rules. People arrive at those four surfaces for different reasons, so different things need predicting.

A second, larger split sits on top. One family of algorithms orders posts from accounts you follow. The other picks content from accounts you do not. Internally the second one is called recommendations, and it is what brings new people in.

SurfaceWhat weighs most
FeedA mix of follows and recommendations. Predicted dwell time, comment, like, send, profile tap
StoriesViewing history, engagement history, estimated closeness to the author
ReelsLikelihood of a send, completion rate, reaction to the audio
ExploreRaw popularity of a post counts considerably more than in Feed

That has a practical consequence. A creator whose reels collapsed while stories carried on as usual did not run into a blanket penalty. Two separate systems simply scored the same account differently.

How a post earns its slot

Mosseri then described the mechanism itself. It is short, and it repeats across every surface.

  1. The platform looks at what the viewer has historically done on it.
  2. From that history it predicts the odds of an action: a like, a comment, a send to a friend, a video view, a watch to the end.
  3. Those predictions combine into a relevancy score for every post the viewer could be shown.
  4. Posts get ordered by that score.

Mosseri called those actions a proxy. Interest cannot be measured directly, so the platform measures whatever resembles it.

Recency carries real weight. It is why the feed looks roughly chronological without being strictly chronological, which is close to his exact wording.

In earlier public statements he put three measures at the top: watch time, likes per reach and sends per reach. The new video assigns no weights, and the order in which he lists actions has not moved.

A relevancy score is recalculated per viewer. There is no single “score” for a post that a creator could look up, in Insights or anywhere else.

The levers you actually control

Ranking is an imperfect science, in his words. That leads to the second half of his point: the platform hands viewers controls for reshaping their own feed. Most go unused because most people never hear about them.

  • Favourites. A separate feed of hand-picked accounts, strictly in time order.
  • The Following feed. Only accounts you follow, with no recommendations at all.
  • Interest topics. The list driving your recommendations is visible and editable.
  • Not Interested on a post. Pushes similar content out of future recommendations.
  • Hidden Words. Filters comments and requests containing words you choose.
  • Why You’re Seeing This. Explains the reason a specific post reached you.

This is exactly where the promised follow-up goes. Mosseri said a dedicated video will cover the controls people have over their own experience.

When reach drops

A creator’s first reaction to falling reach is usually “I have been banned”. The video never mentions bans, yet the explanation answers the worry anyway.

A post’s position is recomputed for every viewer, every time. Something that pulled reach yesterday loses to a fresher post today in front of the same person. That is ordering, not punishment.

Shadowbanning does not exist as an official status. Limited recommendations do exist, and the owner can see that state in Account Status inside the app, where an appeal is filed too.

From the outside, without logging in, one thing is checkable: whether a post appears in hashtag feeds. Popular hashtags are useless for this, since a post slides out of sight in minutes purely on volume. Slow feeds are the telling ones, because a post has no reason to vanish there.

Post visibility across hashtag feeds
Post from 2 Septemberfound in 3 of 4 feeds
Post from 30 Augustfound in 4 of 4 feeds
Post from 28 Augustfound in none
#autumn, #londonexcluded as too fast
What a visibility report looks like. No verdict in it, just findings.

What cannot be known

  • The weight of any individual signal. Instagram publishes no figures, and the video named none.
  • The reason one specific post underperformed.
  • Your own post’s relevancy score, since it is computed per viewer.
  • Who saw a post and scrolled past it.

If reach fell and the cause is unclear, start with something checkable. Sled looks at whether your recent posts appear in slow hashtag feeds and shows where they are missing. No login, and no verdict declaring a shadowban.

Check post visibility
Is it true Instagram has no single algorithm?

Yes, and Mosseri called it the main misconception. Feed, Stories, Reels and Explore run separate ranking systems, and on top of that sits a split between ordering accounts you follow and recommending accounts you do not.

Is the feed chronological?

Roughly, not strictly. Recency weighs enough to keep the order close to time-based. For strict chronology, switch to the Following feed or Favourites.

Can recommendations be reset?

There is no reset button. Recommendations shift through Not Interested on posts, edits to your interest topics, and unfollowing. It accumulates gradually, so nothing rebuilds in one evening.

Does a drop in reach mean a shadowban?

Usually not. Reach depends on how a post competes against fresher ones for a slot in front of each viewer. Limited recommendations do exist as a state, and the owner can see it in Account Status.

Does posting more often help?

Mosseri’s explanation suggests the opposite. He described ranking as the thing that removes the advantage of whoever posts most. More posts means more candidates, but each is scored on its own and viewer attention does not grow.

Sources

Break down the views