The Instagram Reels algorithm in 2026 selects about 100 candidate reels for each viewer, predicts how that viewer will react to each one and orders them by score. For creators it works as a widening test: followers first, then a small group of non-followers, then larger groups while engagement per viewer holds. Watch time is judged against reels of the same length, sends outrank saves, trial reels are ranked separately, and Instagram publishes no reach cap for small accounts.
The Instagram Reels algorithm in 2026 is a ranking system that selects about 100 candidate reels for each viewer, predicts how that viewer will react to each one, and orders them by a relevance score. From the creator's side the same system looks like a widening test: followers typically see a reel first, then a small group of non-followers, then larger groups for as long as each one keeps watching and sending it.
Checked October 2026 against Meta's Reels system card (last updated November 11, 2025), Instagram's February 2025 ranking update, its creator FAQ and ranking explainer, and the Help Center pages on which reels are recommended, trial reels and recommendation eligibility. Instagram publishes signals and predictions, never weights or view thresholds, so this page has none.
This is the ranking deep-dive from our Instagram Reels guide, which covers making and posting a reel. It stays on Reels only; for Feed, Stories, Explore and Search, use the Instagram algorithm guide. Below: the pipeline, every prediction Meta lists, how watch time is judged by length, sends versus saves, the trial reels test and the small-account question.
How does the Reels algorithm work?
Meta calls the system Reels Chaining and describes it in four stages. It runs for one viewer at a time, which is the part most explanations skip: your reel is never ranked in general, only against the other candidates for a particular person at a particular moment.
- 01Gather inventory
Reels from accounts the viewer follows, plus reels similar to ones they engaged with recently.
- 02Select about 100
A lightweight model picks the most relevant, using length, similarity and fit with the viewer's interests.
- 03Filter
Integrity checks remove non-recommendable reels and anything that may break the rules.
- 04Predict
Models estimate how this viewer will react: skip, watch, reshare, comment, follow.
- 05Score and order
A relevance score is calculated for about 100 reels. Higher scores are shown sooner.
Source: Meta Transparency Center, Instagram Reels Chaining, checked October 2026
Instagram's ranking explainer adds the order of the inputs for Reels. First comes the viewer's own activity, then their history with the person who posted, then information about the reel such as its audio, visuals and popularity, and last information about the poster, including follower count and engagement level. Follower count is in the model, at the bottom of the list.
The creator FAQ says the same thing from the other side: the most important things Instagram considers are "how much engagement per viewer the content has received" and "how likely the viewer will be to watch and engage." Per viewer is the phrase to remember. Totals do not rank; rates do.
Reels algorithm explained: every prediction Meta lists
The system card lists nine predictions and the signals behind each. Read the third column as our interpretation; the first two are Meta's.
| Instagram predicts whether you will | Signals Meta lists | What it means for your reel |
|---|---|---|
| Watch less than 3 seconds | Times people watched at least 3 seconds, times it was dismissed, skips within 2 seconds of opening, views with sound on | The opening decides whether the rest is ever scored well. Skip rate in Insights is your view of this. |
| Watch more than 95% of viewers of same-length reels | The length of the reel and how much time people have spent viewing it | Your reel is compared with reels of its own length. |
| Reshare the reel | Time people have spent viewing it; the viewer's past reshares and whose posts they were | Reels that hold attention and resemble what a viewer already sends get predicted as sendable. |
| Share it off Instagram | The viewer's time watching reels and past external shares; how often this reel was shared externally | A link copied to a group chat or another app counts as a signal too. |
| Comment | What and whom the viewer has commented on before, including in the past hour; time viewers from their country spent on the reel | Comments are predicted per viewer; bait does not change who is a commenter. |
| Follow the author | Followers the author gained through this reel; which authors the viewer has followed | A reel that converts viewers into followers is treated as more valuable. |
| Use the audio | How often the viewer taps audio links and "Use audio" | Original or remixable audio gives the system one more thing to predict. |
| Tap "Interested" | Likes on the reel, videos the viewer has saved, profiles they have opened | Likes and saves feed this prediction rather than having one of their own on the card. |
| Open a tall Explore reel full screen | The viewer's Explore and Reels viewing history | Covers decide whether an Explore viewer enters the Reels tab through you. |
Two things stand out. Time spent viewing shows up as a signal behind several predictions, which is why watch time sits in every official summary. And the card has no prediction for a like or a save on its own, even though Adam Mosseri named likes in his top three. The card is a sample of how the system works, not a full specification, so treat it as the floor of what is measured.
How a reel is distributed after you post
The viewer-side pipeline produces a pattern creators can see: reach arrives in rounds. Instagram describes the rounds in its February 2025 update and its creator FAQ.
- 1Eligibility
A reel that is not recommendable can stay up and reach followers, but it is removed from the candidates for everyone else.
- 2Followers first
Instagram says people who follow you are typically the first to see a new reel.
- 3A small group of non-followers
People Instagram predicts will enjoy it, matched on what viewers with similar interests engage with.
- 4Engagement per viewer is read
Watch time, sharing, following, liking, commenting and profile visits, divided by the people who saw it.
- 5Wider groups
The best performers go to a slightly wider audience, then the best of those to a wider one again.
- 6The long tail
Instagram says it can take time for your total potential audience to see a reel, and suggests replying to comments for 7 days.
Source: Instagram for Creators, February 2025 update and creator FAQ, checked October 2026
Instagram publishes no number for any round: not the size of the first group, not a pass mark. A reel stops spreading when each new audience responds less strongly than the candidates it is competing with, which is how most reels end. Our guide to the Reels viral formula turns these rounds into a reel structure, so this page will not repeat it.
Watch time by reel length: the benchmark Instagram actually uses
There is no official watch-time benchmark by reel length, and we found no independent study rigorous enough to quote. What Instagram does publish is the comparison it runs, and it is relative, not absolute.
- Signals listed: the length of the reel and the time people spent viewing it
- So 12 seconds watched means one thing on a 15-second reel and another on a 90-second reel
Source: Meta Transparency Center, Instagram Reels Chaining, checked October 2026
That has a practical consequence: comparing your 20-second reel with your 75-second reel on raw average watch time tells you nothing. Compare each reel with your own reels of similar length. Instagram gives you the inputs: average watch time (watch time divided by initial views), the retention chart added in August 2025, and skip rate, which Instagram defines as "the percentage of views from people who decided to skip your reel during those first 3 seconds."
Build the benchmark yourself. Sort your last 20 to 30 reels into the buckets below and work out watch-through for each: average watch time divided by length. Your median per bucket is the number to beat. As illustrative arithmetic, 11 seconds on a 20-second reel is 55% watch-through and 21 seconds on a 60-second reel is 35%; those two figures are not comparable with each other, only with other reels in the same row.
| Length bucket | Record for each reel | Your benchmark |
|---|---|---|
| Under 15 seconds | Average watch time, skip rate, replays implied by watch time above length | Your median watch-through for reels under 15 seconds |
| 15 to 30 seconds | Average watch time, skip rate, the second where the retention line bends | Your median for 15 to 30 seconds |
| 30 to 60 seconds | Average watch time, skip rate, retention at the halfway mark | Your median for 30 to 60 seconds |
| 60 to 90 seconds | Average watch time, skip rate, retention at each section change | Your median for 60 to 90 seconds |
| 90 seconds to 3 minutes | Average watch time, skip rate, retention at each section change | Your median for 90 seconds to 3 minutes |
Watch-through is our working ratio, not an Instagram metric. The rule it protects you from is real, though: a reel over 3 minutes is not recommended to new audiences, whatever its retention. Loops are a separate question, covered in do looped reels get more views.
Sends vs saves: which counts more for Reels?
Sends. Mosseri's January 2025 list was "watch time, likes and sends", and he added that "sends are slightly more important for unconnected content", meaning viewers who do not follow you. The system card agrees in its own way: it predicts resharing and sharing off Instagram, and has no prediction for saving.
- Named in Mosseri's top three signals
- Slightly more important with non-followers
- Two predictions on the Reels card: reshare and share off Instagram
- Puts the reel in front of a new person directly
- Not in Mosseri's top three
- No prediction of its own on the Reels card
- Saved videos feed the "Interested" prediction
- Named for Explore, alongside likes and shares
Nobody outside Instagram knows the ratio between them, and any "one send equals so many likes" figure is invented. The working rule is simpler: a save says the reel was useful to one person, a send says it was worth someone else's time. On a surface built for strangers, the second is the stronger evidence. Our comparison of saves, likes and shares covers the other surfaces.
The trial reels distribution test
A trial reel is the non-follower round on its own, with your followers taken out. That makes it the cleanest test of whether an idea works for strangers.
- Who sees it. Accounts that do not follow you. It stays out of your followers' feeds and off your profile until you share it to everyone.
- How it is ranked. The Help Center says trial reels "are evaluated independently in Instagram's ranking and don't impact the performance or ranking of your standard reels content."
- What you get back. Views, likes, comments and shares within 24 hours. Instagram notes trials "may take longer to get views" because nobody who follows you is in the audience.
- Who can use it. A public account with at least 200 followers (professional) or 1,000 (personal). An account that is not eligible for recommendations loses access.
- The catch. It must be a new reel. A trial "may get limited reach if Instagram detects that you've previously shared the same content."
The matrix is our reading, not Instagram's. The full testing loop is in the trial reels strategy guide, and trial reels vs regular reels covers how to compare the two fairly. Instagram Ignited builds this into a weekly test-and-review routine, with the Insights sheet and the hook variations to run each week.
Is there a distribution cap for small accounts?
Instagram publishes none, and everything it has said points the other way. The February 2025 update says reels are distributed on engagement, "not just the creator's follower count." In September 2026 Mosseri described "breaking" content: showing public posts from smaller creators to non-followers to see how people respond.
What small accounts feel as a cap is usually one of three things:
- A small first audience. Followers typically see a reel first. With few followers, the first read is a handful of people, so results swing more from reel to reel. That is noise, not a ceiling.
- Eligibility. If the account or the reel is not recommendable, nothing reaches non-followers. That is a switch, not a slope, and Account Status shows it.
- Competition per viewer. Your reel has to make one viewer's shortlist of about 100 and then outscore the rest. A bigger account does not get a bigger shortlist.
These are the numbers Instagram and Meta have actually published about Reels distribution. Anything else you see quoted, including view thresholds for "stages", is someone's estimate.
| Number | What it is | Source (checked October 2026) |
|---|---|---|
| About 100 | Reels selected and scored for a viewer in one ranking pass | Meta Reels system card |
| 3 seconds | The watch threshold behind the skip prediction and the skip rate metric | Meta Reels system card; Insights |
| 95% | The share of same-length viewers Instagram predicts you will out-watch | Meta Reels system card |
| 3 minutes | The longest reel Instagram recommends to people who do not follow you | Help Center |
| 200 or 1,000 followers | Minimum for trial reels: professional or personal account | Help Center |
| 24 hours | Trial reel insights appear within this window | Help Center |
| 10 or more reels a month | Average among creators with the greatest net follower growth | Instagram creator FAQ |
| 7 days | The window in which Instagram suggests replying to comments on a reel | Instagram creator FAQ |
| 30 days, rolling | The period over which most posts must be original to regain recommendations | @creators, April 2026 |
Instagram Reels algorithm 2026: what changed and when
The models change continuously; Instagram announces the rules and tools around them. These are the Reels-relevant ones.
- Dec 2024Trial reels launch
Test a reel with non-followers before your followers see it.
- Jan 2025Top three signals named
Mosseri: watch time, likes and sends.
- Feb 2025Ranking update described
A small non-follower audience first, then wider groups for the best performers.
- Aug 2025Retention chart and skip rate
Insights show where viewers drop, and skip rate replaces view rate.
- Dec 2025Your Algorithm for Reels
Viewers see and edit the topics shaping their Reels tab; extended to Explore and Feed in 2026.
- Apr 2026Originality rule covers all formats
Accounts that mainly repost lose recommendations until most of a rolling 30 days is original.
- Sep 2026Four principles restated
Mosseri: originality, recency, breaking smaller creators and larger models.
Source: Instagram and Meta announcements, checked October 2026
Sources: trial reels, Your Algorithm and the originality update. Viewer topic controls matter to creators for one reason: a reel that is clearly about one recognisable topic is easier to match to the people who asked for that topic.
Mistakes that stall distribution
- Judging totals instead of rates. Instagram ranks on engagement per viewer. A reel with fewer views and a higher send rate is the better reel.
- Comparing reels of different lengths. The system does not, so you should not either.
- Re-uploading the same reel. Instagram's ranking explainer lists reels already posted on Instagram among those that are less visible. Fix the hook and post a new cut instead.
- Leaving another app's watermark on. The Help Center lists borders, logos and watermarks among reels shown less often, along with muted, blurry or low-resolution video.
- Changing topic every week. Audience matching works from what similar viewers engaged with. Instagram's FAQ says content on a theme that earned distribution before "could get similar distribution."
- Going quiet. Instagram reports that creators with the greatest net follower growth post, on average, 10 or more reels a month.
When reach falls on every reel at once, stop adjusting content and diagnose. The free Instagram Reach Drop Diagnoser ranks the likely causes from your own Insights numbers, and why reels stop getting views pairs each cause with the number that confirms it. If the account had a baseline and lost it, use the 7-step recovery for a sudden drop.
What to do with this each week
- Check Account Status once: recommendations are either on or off
- Log skip rate and average watch time for every reel you post
- File each reel in its length bucket and compare it with your own median
- Look at the retention chart and note the second where the line bends
- Work out sends per account reached, and write down which reel won
- Post one new hook as a trial reel and read it after 24 hours
- Reply to comments on last week's reels while they are inside 7 days
- Keep every reel original, watermark-free and at 3 minutes or less
None of this promises a view count. It does make sure the next reel is judged on the idea, not on an avoidable mistake.
Instagram Reels algorithm FAQ
How does the Instagram Reels algorithm work in 2026?
For each viewer, Instagram gathers candidate reels, uses a lightweight model to select about 100 of the most relevant, removes anything that is not recommendable, predicts how the viewer will react to each one and orders them by a relevance score. For a creator that looks like a widening test: followers typically see a reel first, then a small group of non-followers, then larger groups while engagement per viewer holds up.
What is the most important signal in the Reels algorithm?
Instagram names no single one and publishes no weights. Adam Mosseri said in January 2025 that the top three signals are watch time, likes and sends, with sends slightly more important for people who do not follow you. Instagram's creator FAQ says distribution depends most on engagement per viewer: watch time, sharing, following, liking, commenting and profile visits. Watch time appears in every version of the answer.
Does the Reels algorithm favor short reels?
Not by rule. Meta's Reels system card says Instagram predicts whether you will watch more of a reel than 95% of people who watched reels of the same length, so a reel is compared with others of its own length, not with shorter ones. The only length rule is the ceiling: reels over 3 minutes are not recommended to new audiences. Make the reel as long as the idea holds attention.
Do saves or sends matter more on Reels?
Sends. Mosseri named sends, not saves, in his top three signals, and Meta's Reels card lists predictions for resharing a reel and for sharing it off Instagram but none for saving it. Saved videos appear only as an input to another prediction. Saves still tell you a reel was useful, and Instagram names them among the key actions for Explore, so track both but build Reels to be sent.
Do small accounts get less reach on Reels?
Instagram publishes no reach cap by follower count. It says reels are distributed on engagement, not just follower count, and Mosseri said in September 2026 that Instagram shows posts from smaller creators to non-followers to see how they respond. Follower count is listed last among the signal groups for Reels. In practice a small account has a smaller first audience, so early results are noisier, not capped.
Do trial reels affect how my regular reels are ranked?
No. Instagram's Help Center says trial reels are evaluated independently in ranking and do not impact the performance or ranking of your standard reels. They are shown to people who do not follow you, may take longer to collect views, and need a public account with 200 followers on a professional account or 1,000 on a personal one. Checked October 2026.
How often does the Instagram Reels algorithm change?
Continuously, without version numbers. Meta's Reels system card was last updated on November 11, 2025, and Instagram announces rule and feature changes rather than model changes: trial reels in December 2024, the retention chart and skip rate in August 2025, viewer topic controls in December 2025 and the wider originality rule in April 2026. Recheck the official pages before trusting any dated advice, including this page.
You know how Reels are ranked. Now run the weekly loop.
Instagram Ignited, included in All Access, turns these signals into a weekly routine: hooks and retention, trial reels, Instagram SEO and the Insights review. One subscription also covers the other three programs, live coaching and the private community.
Reach dropped? Run the numbers first
The free Reach Drop Diagnoser ranks the likely causes from your own Insights, and the free Telegram channel posts Reels rule changes as they land.