GuideFormats & Craft20 min read

Carousels vs reels vs static: the format showdown, settled with mechanics

Instagram's Feed card documents a carousel-specific prediction that does not exist for a static post. Here is what Meta's own ranking documents, and two studies covering millions of posts, say about which format to reach for.

Aubrium ResearchEditorial · 8 February 2026

You made a carousel because someone said carousels work. You made a reel because someone else said reels win. The carousel did fine. The reel did better. Now you are trying to reverse-engineer why, and every video on the subject gives you a different answer.

So you go looking for something solid. A claim you can check. Most of what you find is someone's month of posts, or a trend observation, or a strong opinion shaped like a fact. The studies that do exist count millions of posts and show reels ahead. But they rarely explain why. And none of them open the machine to show you what it is actually measuring.

We went looking for the mechanics. Not which format got more reach last quarter. Which format the software is built to measure differently. Meta publishes pages explaining how each part of Instagram decides what to show you. It calls them system cards. We read four of them. We also read what Adam Mosseri wrote about ranking. He runs Instagram. And we read two engineering posts from the people who build the system.

Here is what the documents say. Instagram's Feed card names a prediction about carousels that does not exist for a static post. It guesses how likely you are to swipe through the whole carousel. That guess has its own signals. How many cards you view. How many times you swipe to the next one. Whether the carousel was reordered in your feed. The guess exists because a carousel is not one thing. It is a set of cards. And the software wants to know if you will bother to see them all.

A static post has no such prediction. Neither does a reel, on the Feed card. They get scored on other guesses instead. How likely you are to skip them. How much time you will spend. Whether you will tap through to the profile. The bundle of guesses is different for each format. And no two parts of Instagram use the same bundle.

That is the documented mechanical difference. It does not tell you which format wins. It tells you what the machine is trying to predict, format by format. What follows is a reading of what Meta publishes, and the largest disclosed-methodology format studies we could verify. Every source is linked, dated and quoted. One section reports a finding that cuts against what you would expect if carousels had a ranking edge.

The predictions are not the same across formats

Start with what a prediction is. The software does not know if you will like a post. It guesses. That guess is a prediction. Every part of Instagram makes a different set of guesses. Your feed makes some. The Explore tab makes others. The Reels tab makes a third set. Meta calls each part a surface. Each surface has its own list.

The Feed system card was updated 29 June 2026. It names nine predictions. We list them in the order Meta does:

How likely you are to enjoy this post
How likely you are to skip this post
How much time you're likely to spend on the author's profile after seeing this post
How likely you are to share a post with someone in a direct message
How much time you're likely to spend on this post
How likely you are to swipe through the whole carousel
How likely you are to skip over the first post in feed
How likely you are to spend more than 10 seconds on the first post
How likely the first post you see is to keep you spending time in your Feed during this session

Number six is the one to hold on to. It exists only for carousels. Meta's card says so in the signals that feed it. Those signals include "How many of the carousel's cards you view" and "How many times you swipe to the next card". A static post has no cards to swipe through. So that prediction does not run.

The other eight predictions apply to all formats. Or they apply to none, because the format does not fit. Prediction seven and eight are about the first post in your feed. That post can be any format. Prediction nine measures whether that first post keeps you scrolling. Again, any format qualifies.

So a carousel gets scored on one thing a static post does not. That does not mean it wins. It means the software is asking a different question. For a carousel, part of the score comes from whether you will swipe through it. For a static post, that question never gets asked.

Now look at Reels. The Reels Chaining card was updated 11 November 2025. Chaining means the run of reels that keeps coming once you open one. Its predictions are all about reels. Nine of them:

How likely you are to use audio from a reel in one you create
How likely you are to watch less than three seconds of a reel
How likely you are to click Interested if that option is presented
How likely you are to comment on a reel
How likely you are to watch more of a reel than 95% of users who watched reels of the same length
How likely you are to click tall, rectangular reels in Explore to view them full screen in the Reels tab
How likely you are to reshare a reel
How likely you are to share a reel off Instagram
How likely you are to follow the author of a reel

Prediction two is the one to notice. The card says a signal feeding it is "How many times the post has been skipped within two seconds of being opened." A reel can be skipped in two seconds. A carousel cannot be opened and skipped that fast. The format does not work that way. The question makes no sense for a carousel.

So the bundle of guesses changes by format. And it changes by surface. A reel in your feed gets scored by the Feed card's predictions. A reel in the Reels tab gets scored by the Reels Chaining predictions. Same format. Different part of the app. Different list of guesses. There is no universal answer to which format ranks better. The software is not even asking the same questions.

Feed Recommendations scores informativeness with an AI

Your feed shows two kinds of post. Posts from accounts you follow. And posts from accounts you do not. The Feed card covers the first kind. The Feed Recommendations card covers the second. That card was updated 29 June 2026. It lists ten predictions.

One of them is not about you. It is about the post. "How informative a post is." The card says how that guess gets made. In Meta's own words:

This signal is purely generated by LLM to assess the post's content quality — Meta Transparency Center, Instagram Feed Recommendations AI system card

LLM is short for large language model. It is software trained to read and write text. An AI reads the post. It scores how informative the post is. That score feeds a ranking guess. Every other prediction across the four cards we read guesses something about a viewer. This one scores the content itself.

The card does not say which formats that applies to. It does not say a carousel scores higher on informativeness than a reel. Or that a static image scores lower. It names the prediction. It names the method. An LLM reads it. That is all it says.

But the implication sits there. A carousel holds more cards. Each card can carry text and an image. More room for information. Whether the AI scores that higher is not published. The mechanic is. An AI is reading the content and judging it. For posts you do not follow. In your feed.

Mosseri says the system adapts to your format preference

Adam Mosseri runs Instagram. He wrote about ranking on 31 May 2023. The post is called Instagram Ranking Explained. It walks through Feed, Stories, Explore and Reels. One sentence in the Feed section addresses format directly:

We've also started considering other factors like format, so if we notice you prefer photos, we'll show you more photos. — Adam Mosseri, Instagram Ranking Explained, 31 May 2023

That is a claim about personalization. Not about one format ranking higher than another for everyone. The system watches what you engage with. If you prefer photos, you see more photos. If you prefer reels, you see more reels. The algorithm adapts to you.

So the answer to which format wins is not the same for every person. Two people open Instagram. One watches reels to the end. The other swipes through carousels. Instagram will show each of them more of what they already do. That is what the head of Instagram says the system does.

That claim cuts against a universal ranking. If reels always ranked higher for everybody, personalizing by format would not matter. Everyone would already be seeing reels. The fact that Instagram tunes the feed to your format preference means the preference varies. And the system responds to it.

The Fanpagekarma study: reels lead on reach, over 3 million posts

The largest format study we could verify with a disclosed method comes from Fanpagekarma. It ran from January to April 2026. It analyzed more than 3 million Instagram posts. Reels, carousels and single images. The study split accounts into seven size bands. From under 10,000 followers to over 1 million.

The finding is consistent. Reels get the highest reach per follower in every band. Small accounts under 10,000 followers see reels reach 4.9% of followers per post. Carousels reach 3.2%. Single images reach 3.5%. As accounts grow, all three formats lose reach per follower. But reels stay on top. Accounts over 1 million followers see reels reach 2.3% per post. Carousels reach 1.2%. Single images fall to 0.39%.

The study also measured non-follower reach. Reels lead there too. By roughly two to three times the reach of carousels. In every size band. That is the discoverability advantage. A reel reaches more people who do not follow you than a carousel or a static image does.

But reach is not engagement. The same study measured interactions per post. For very small accounts, single images win. They get 0.51% interactions per follower. Reels get 0.43%. Carousels get 0.38%. That flips as the account grows. In the middle bands, 100,000 to 1 million followers, reels overtake images on interactions. Carousels stay in third.

So reels reach more people. But for the smallest accounts, static images get more interactions per person reached. The pattern Fanpagekarma found is not one format wins. It is reels win on reach, images sometimes win on depth, and carousels sit between them.

The study's method is disclosed. Over 3 million posts. Four months. Seven account-size bands. Median reach per follower and per post. That is enough to check. The weakness: it measures outcomes, not mechanics. It tells you what happened in those four months. It does not tell you why Instagram's ranking models produced that result. For that, you need the system cards. And the cards do not publish the weights.

Buffer's 2026 report: carousels lead on engagement rate, reels on reach

Buffer published a report in March 2026. It analyzed over 45 million posts from Buffer accounts. The finding splits the same way. Reels get the most reach. Carousels get the highest engagement per person reached.

Buffer measured engagement rate as a percentage of reach. Carousels: 6.9% median engagement rate. Single images: 4.4%. Reels: 3.3%. So a carousel earns more engagement from the people who see it than a reel does. By 109%, in Buffer's wording.

But Buffer also ran a separate analysis. Four million posts, from January 2022 to October 2024. Reels got 1.36 times the reach of carousels. And 2.25 times the reach of single images. So reels reach more people. Carousels engage those people more deeply. Buffer's conclusion, verbatim: "The 'best format on Instagram' has no single answer — it really depends on your goals."

Both studies agree on the split. Reels maximize reach and discovery. Carousels maximize engagement depth. Neither is wrong. They measure different things. And Instagram's own system cards document the mechanical reason. A carousel gets scored on whether you swipe through the whole thing. A reel gets scored on whether you skip it in two seconds. The format shapes what the software predicts. The predictions shape the score. The score shapes the reach.

What the studies measure, and what they cannot

Both Fanpagekarma and Buffer disclose their methods. Sample sizes in the millions. Time windows of months. Median reach and engagement, broken out by account size or calculated as rates. That is more rigorous than most format advice. It is also not an experiment.

An experiment would show the same account post the same content in three formats. To the same audience. At the same time. With the three formats randomly assigned. Then measure which one Instagram showed to more people. That is the design that would isolate format as the cause.

Neither study does that. They measure what actually happened across millions of posts. From different accounts. With different content. Over different months. The posts that were made as reels are not the same posts that were made as carousels. So the format difference is mixed with a content difference, an audience difference, and a time difference.

That does not make the studies useless. It makes them observational. They tell you what the pattern was, in the real world, over the period they measured. They do not prove that changing your next post from a carousel to a reel will increase its reach. The confounds are too large.

The pattern is still useful. Reels reach more people, in both studies, across every account size. That is evidence. It is not proof of what would happen to your account. But it is the best large-scale evidence that exists with a disclosed method.

One more limit. Both studies measure the past. Fanpagekarma ran January to April 2026. Buffer's reach analysis ran January 2022 to October 2024. Instagram's ranking models change. Meta does not publish when. The system cards get updated with no changelog. So a study from 2024 measures a system that may not exist anymore. The result tells you what was true then. Not what is true now.

The one finding that cuts against carousel ranking

Here is the inconvenient part. If carousels had a ranking advantage, you would expect them to get more reach than static images. They get scored on an extra prediction that static images do not. How likely you are to swipe through the whole carousel. That prediction has its own weight. It feeds the final score.

So carousels should outperform static images on reach. If that extra prediction carries meaningful weight.

The studies do not show that. Fanpagekarma found carousels reach 3.2% of followers for small accounts. Static images reach 3.5%. Static images win. For accounts over 1 million followers, carousels reach 1.2%. Static images reach 0.39%. There, carousels win. But the difference flips depending on account size.

Buffer's study does not break out static images separately from carousels in the reach analysis. It groups them together. So we cannot check Buffer's data against Fanpagekarma's on this split. But Buffer does report engagement rate. Carousels win on engagement per person reached. Static images come second. Reels come third. That is consistent with carousels holding attention longer. But holding attention is not the same as getting reach.

If the carousel-specific prediction carried a large positive weight, carousels should reach more people than static images do. Not just engage them more deeply. The evidence for that is mixed. And that is the finding that cuts against expecting carousels to have a reach advantage. The mechanical difference exists. The documented outcome difference does not point one way.

Which format to reach for, by what you are trying to do

The system cards tell you what gets measured. The studies tell you what tends to happen. Put them together and the advice becomes concrete.

Reach for a reel when you want discovery. Reels get scored on watch time and whether you skip them in two seconds. They reach more people who do not follow you. They maximize the chance that someone new sees your work. That is consistent across both large studies. Every account size. Every time period measured.

Reach for a carousel when you want depth. Carousels get scored on whether you swipe through the whole thing. They hold more information per post. An AI scores how informative they are, for posts from accounts you do not follow. They earn higher engagement per person reached. That suits teaching something. Or telling a story across cards. Or holding someone's attention long enough to make a point.

Reach for a static image when your account is small and you want interactions. The Fanpagekarma study found static images get the most interactions per follower for accounts under 10,000 people. That advantage disappears as the account grows. But for a new account, a single strong image can outperform a reel or a carousel on getting people to comment, like or save.

None of that is a law. Instagram personalizes by your format preference. If your audience watches reels to the end, they will see more reels. If they swipe through carousels, they will see more carousels. Mosseri said so. So the format that works for one account may not work for another. The system adapts to the viewer, not to a universal rule.

But the mechanical differences are real. A carousel gets scored on something a static post does not. A reel gets scored on something a carousel cannot. The bundle of predictions changes by format. And by surface. The advice that one format always wins is too flat to be true. The machine is asking different questions. Answer the one that matches what you are trying to do.

What we could not determine

Four things we looked for and could not verify.

The weights on any prediction. Meta publishes the list of predictions. It publishes the signals that feed them. It does not publish the weight each prediction carries in the final score. Not on any surface. Not for any format. So every claim about which prediction matters most is unproven. Including the claims in this post. We report what gets measured. We do not know how much each measurement counts.

Whether the carousel swipe-through prediction carries a positive or negative weight. The Feed card predicts how likely you are to swipe through the whole carousel. A high likelihood could add to the score. Or it could subtract. The card does not say. The engineering posts do not say. We assume it adds, because a person swiping through all the cards sounds like engagement. But that is our reasoning. Not a published fact.

Any controlled experiment on Instagram showing format causes reach. The two large studies we cite are observational. They measure millions of posts. They find reels get more reach. That is evidence, not proof. A controlled experiment would post the same content in three formats and measure which Instagram shows to more people. No such experiment is published with a disclosed method that we could find. So the causal claim that switching your next post to a reel will increase its reach has no experimental proof. Not a weak experiment. None.

How the Reels tab ranks a carousel. The Reels Chaining system card is about reels. A carousel is not a reel. Can you even post a carousel to the Reels tab? We do not know. The card does not say. If you can, we do not know how it gets scored. The predictions are all written for reels. Whether a carousel gets forced into that scoring, or skipped, or scored by a different model, is not in any document we read.

Where this leaves you

You made a carousel. You made a reel. One did better than the other. Now you know why the machine might have treated them differently. Not because one format is universally favored. Because the software asks different questions of each format. And the people watching your posts answer those questions with their behavior.

Instagram guesses whether you will swipe through a carousel. It guesses whether you will skip a reel in two seconds. It guesses how much time you will spend on a post. It guesses if you will share it, save it, or tap through to the profile. Every format gets scored on a different bundle of guesses. So the format that maximizes one guess may not maximize another.

Reels reach more people. That is documented in two large studies. Carousels engage people more deeply. That is documented in the same two studies. Static images win on interactions for very small accounts, in one study. And Instagram personalizes to your format preference. That is documented by Mosseri.

So the question is not which format wins. It is which format suits what you are trying to do. And whether your audience already has a format preference strong enough to override the general pattern. You will only know that by checking your own numbers. Not someone else's study. Yours.

Someone will tell you the algorithm favors reels. They are half right. Reels reach more people, in the largest studies that exist. But reach is not the only thing the algorithm measures. It also measures depth, time, and whether you bothered to look at all the cards. Those favor carousels. The algorithm is not one thing with one preference. It is over a thousand models making different guesses on different surfaces. Some of those guesses suit reels. Some suit carousels. Which matters more depends on what you need.

How we did this

Scope: this post reads four Instagram system cards, two Meta engineering posts, one post by Adam Mosseri, and two large-scale format studies with disclosed methods. No data collection. No experiment of our own.

How we read them: the four system cards, Mosseri's post, and the two engineering posts were opened directly on 17 August 2026. Every quoted sentence was copied as a reader sees it. Six of those sources are built by code in the browser and send an empty page to a fetcher. So they were read in a browser. The two studies were read at their published URLs on 18 August 2026. The Fanpagekarma study is dated 2026 and reports data from January to April 2026. The Buffer study is dated 19 March 2026 and reports data from two windows: 45 million posts for engagement rate, and over 4 million posts from January 2022 to October 2024 for reach.

Limits: seven of nine sources are first-party statements. Meta, Instagram, or Buffer describing their own systems or their own data. A system card is a description Meta chose to publish. Not an audited account. Meta rewrites these cards without saying when. So every quotation carries the card's updated stamp. The two studies are observational, not experimental. They report what happened across millions of posts. They do not prove that format caused the reach difference. Both studies rely on their own customer data. Neither is peer-reviewed.

Sources

Instagram Feed AI system card, stamped UPDATED JUN 29, 2026. Meta Transparency Center. Source of the nine Feed predictions and the carousel swipe-through prediction. Read in a browser. Retrieved 17 August 2026.

Instagram Reels Chaining AI system card, stamped UPDATED NOV 11, 2025. Meta Transparency Center. Source of the nine Reels Chaining predictions and the two-second skip signal. Read in a browser. Retrieved 17 August 2026.

Instagram Feed Recommendations AI system card, stamped UPDATED JUN 29, 2026. Meta Transparency Center. Source of the LLM-generated informativeness prediction. Read in a browser. Retrieved 17 August 2026.

Instagram Explore AI system card, stamped UPDATED JUN 22, 2026. Meta Transparency Center. Read in a browser. Retrieved 17 August 2026.

Adam Mosseri, Instagram Ranking Explained, 31 May 2023. About Instagram. Source of the format personalization quote. Retrieved 17 August 2026.

Journey to 1000 models: Scaling Instagram's recommendation system, 21 May 2025. Engineering at Meta. Retrieved 17 August 2026.

Scaling the Instagram Explore recommendations system, 9 August 2023. Engineering at Meta. Retrieved 17 August 2026.

Fanpagekarma, Instagram Format Reach Analysis, reporting data from January to April 2026. Sample: more than 3 million reels, carousels and image posts. Seven account size bands from under 10,000 to over 1 million followers. Method: median reach per follower and post. Retrieved 18 August 2026.

Buffer, Data Report: The Best Content Format on Social Media in 2026, published 19 March 2026. Sample: over 45 million posts for engagement rate analysis. Over 4 million posts from January 2022 to October 2024 for reach analysis. Method: median engagement rate as percentage of reach. Reach measured as multiples. Retrieved 18 August 2026.

Changelog

2026-08-18: First version drafted.

carouselsreelsinstagramrankingreach