The Instagram Algorithm in 2026: What Actually Determines Your Reach

Updated September 18, 2026.

If your Instagram reach dropped, the useful question isn't “What does the algorithm want today?” Instagram doesn't run one universal ranking formula. Feed, Stories, Explore, Reels, Search, and recommendations use different systems to predict what each person is most likely to watch, like, send, save, reply to, or otherwise find relevant.

For creators, the practical picture is simpler. Instagram has repeatedly pointed creators toward three metrics when diagnosing performance: watch time, likes per reach, and sends per reach. Then there is another layer that matters before ranking even starts: whether your content is eligible to be recommended to people who don't follow you.

This guide separates those two problems and explains what actually affects Instagram reach in 2026.

There isn't one Instagram algorithm in 2026

Instagram has been explicit about this for years. Different parts of the app are used for different reasons, so they are ranked differently.

You usually open Stories expecting to see people you already know or follow. Explore has the opposite job: it needs to find content from accounts you may never have seen before. Reels also leans heavily into discovery, while the main Feed mixes followed accounts with recommendations.

That matters because a tactic that improves Story placement may do very little for Explore distribution.

Meta's public ranking documentation describes separate AI systems for Instagram surfaces rather than one master score. The company also says these systems use many signals to make predictions about what an individual user is likely to find relevant.

So when someone says “Instagram now prioritizes X,” I would immediately ask: where? Feed? Stories? Reels? Recommended posts?

There are a few signals that are useful across formats, though. Those are the best place to start.

The three Instagram ranking metrics worth watching first

In a 2025 explanation of Instagram ranking, Adam Mosseri highlighted watch time, likes, and sends as the three metrics creators should pay particularly close attention to.

The more useful detail is how he suggested measuring them. Raw numbers can be misleading. A post reaching 500,000 people should naturally collect more likes than one reaching 8,000. The better diagnostic is the response relative to the audience that actually received the post.

1. Watch time

For video, watch time tells Instagram whether people stayed after the Reel started.

This doesn't mean every Reel should be seven seconds long, and it doesn't mean completion rate is the only thing that matters. A longer Reel can have a lower completion percentage and still generate meaningful watch time if people stay for most of it.

Instead, compare similar content against similar content. If two 30-second tutorials cover comparable topics and one consistently loses viewers in the opening seconds, the problem is probably visible before you start blaming hashtags or posting time.

Look at the first seconds, the point where retention falls sharply, unnecessary intros, repeated information, and whether the payoff arrives too late.

2. Likes per reach

Mosseri has said likes are slightly more important for what Instagram calls connected reach: distribution to people who already follow or know your account.

The useful formula is:

Likes per reach = likes ÷ accounts reached × 100

A post with 2,000 likes isn't automatically stronger than one with 600. If the first reached 200,000 accounts and the second reached 8,000, they produced very different viewer responses.

This is also why comparing raw like totals across creators is often a bad way to reverse-engineer Instagram performance.

3. Sends per reach

Sends matter particularly when the goal is reaching people who don't already follow you.

A send happens when someone decides the post is relevant enough to pass to another person, usually through a direct message. That gives Instagram information that a post may travel beyond the audience that already knows the creator.

The practical metric is:

Sends per reach = sends ÷ accounts reached × 100

You don't need to turn every caption into “SEND THIS TO A FRIEND.” Build posts that naturally create that action. A restaurant list someone sends to their partner, a painfully accurate joke sent to a coworker, a useful checklist shared with a client, or a Reel that reminds someone of a friend all have a reason to move through DMs.

Connected reach and unconnected reach are different problems

This distinction explains a lot of confusing Instagram analytics.

Connected reach is attention from people who already have a relationship with your account. Unconnected reach comes from people who don't follow you yet, often through recommendations, Explore, or Reels discovery.

You can perform well with one group and poorly with the other.

Imagine a local bakery with 18,000 followers. Its regular customers may like almost every behind-the-scenes photo. Those posts can perform well with followers because the relationship already exists. A non-follower in another part of the city has no reason to care about an ordinary photo of Tuesday's croissants.

Now the bakery publishes a Reel showing a strange but useful trick for reheating a croissant without drying it out. Existing followers may like it, but people can also send it to someone who buys pastries, save it for later, or watch the demonstration to the end. That content has a much clearer route into unconnected reach.

The account didn't become “more favored by the algorithm.” The post generated signals that fit a different distribution job.

Reach is a two-stage problem: eligibility first, ranking second

Creators often jump straight to engagement metrics. That skips an important question.

Before Instagram decides how widely to recommend a post, the post and account need to be eligible for recommendations.

A piece of content can be allowed to remain on Instagram while still receiving less or no recommendation distribution. Community Guidelines and recommendation eligibility aren't exactly the same thing.

For professional accounts, checking Account Status should therefore be part of any reach investigation. If Instagram tells you that content or the account isn't eligible for recommendations, changing the first three seconds of a Reel won't solve that particular problem.

Original content became an even bigger reach issue in 2026

Originality is one of the clearest changes worth paying attention to this year.

Meta reported in January 2026 that the prevalence of original content in Instagram recommendations in the United States increased by 10 percentage points during Q4 2025, reaching 75% of recommendations.

That doesn't mean Instagram gives every original post a reach bonus. It tells us something more useful: the recommendation system has been deliberately moving toward original material.

Instagram had already been working to replace identical reposts with the original version and reduce recommendation exposure for accounts built around repeated reposting. In April 2026, that approach was expanded further to cover photos and carousels from aggregator-style accounts, not only Reels.

The distinction is about creative contribution. Taking another creator's post, adding a border, changing the playback speed, placing a watermark on it, or uploading a screenshot with credit doesn't necessarily turn the work into original content.

Using existing material as part of genuine commentary, analysis, parody, reporting, or a meaningful creative transformation is a different case.

For brands, this is easy to overlook. A social media calendar filled with reposted customer photos, screenshots, supplier assets, viral memes, and lightly edited third-party clips may look active while producing very little material Instagram has a reason to recommend as original work.

How Feed, Stories, Explore, and Reels rank differently

Surface	What Instagram is trying to predict	What matters in practice
Feed	Which followed and recommended posts a person is most likely to find relevant and interact with	Previous behavior, interest in similar content, expected interactions, time spent, relationship and post information
Stories	Which accounts' Stories a person is most likely to want near the front of the tray	Viewing history, engagement history and the viewer's relationship with the account
Explore	Which unfamiliar posts match the person's interests	Behavior on similar content, information about the post, engagement around the post and predicted interest
Reels	Which videos are likely to keep the person watching and produce a positive response	Watch behavior, likes, sends, previous Reel activity, information about the Reel and recommendation eligibility

The table is deliberately broad. Instagram's ranking systems use far more inputs than a creator can realistically optimize one by one.

That's also why “the algorithm likes carousels” or “the algorithm hates photos” is usually too simplistic. Format changes the behavior available to the viewer. A carousel can create additional time on post and swipes. A Reel creates watch-time data. A Story can generate replies and repeated viewing habits.

The format itself isn't a substitute for making something people want to consume.

What LikeWave's 137-account study tells us about follower count

We wanted to see how much useful information visible follower and like numbers actually give a marketer, so in August 2026 LikeWave analyzed 137 public Instagram accounts across fitness, beauty, travel, food, fashion, gaming, and tech.

Each account's report covered its latest 12 posts, giving us 1,644 recent-post observations.

The median account averaged 2,800 likes per post. The arithmetic mean was 52,610.

That enormous gap matters. A relatively small number of very large or unusually viral accounts pulled the average upward, making a simple “average Instagram likes” number almost useless without context.

Follower count and absolute average likes had a moderate positive Spearman correlation of 0.411 in the sample. Meanwhile, follower count and follower-based like rate moved in the opposite direction, with a correlation of -0.367.

You can read the complete methodology in our Instagram Likes Statistics 2026 study.

There is an important limitation here. Public account data cannot tell us Instagram's internal ranking weights. Our dataset doesn't contain private sends, watch time, unique reach, or Meta's recommendation scores.

What it does show is why follower count and raw likes are weak tools for diagnosing reach on their own. Two accounts of similar size can produce radically different recent engagement patterns.

If you're reviewing a public competitor, LikeWave's free Instagram Likes Viewer can help you compare recent likes, posting patterns, followers, and public engagement. For your own account, Instagram Insights should remain the primary source because it contains private metrics such as reach and content interactions that an outside tool cannot see.

Views matter, but a view doesn't tell you why the post spread

Instagram moved toward Views as the common headline metric across Reels, Stories, photos, and carousels. That makes cross-format reporting easier, but views shouldn't be confused with unique reach.

The same person can generate more than one view.

For diagnosis, I would read views together with Accounts Reached and then ask what happened after exposure. Did people keep watching? Like? Send? Save? Visit the profile? Follow?

A Reel with a large view count and almost no profile action may be doing an excellent job as entertainment while doing very little for your business goal. That's still useful information. It tells you what the content is actually accomplishing.

Saves and comments still matter

You may have seen posts claiming that saves are “the most powerful Instagram signal” or that comments now matter less.

Instagram's public explanations don't support a universal ranking like that.

Saves, comments, profile visits, replies, previous interactions, content information, and many other signals can appear in different ranking systems and predictions. Mosseri highlighting watch time, likes, and sends gives creators a practical shortlist. It doesn't mean every other interaction has been deleted from ranking.

A save is especially meaningful for content designed to be revisited: recipes, workouts, travel plans, checklists, tutorials, product references, and research.

A comment can show interest or start a relationship that affects future behavior. Story replies are even more obviously connected to an ongoing relationship between two accounts.

Measure the interaction that fits the content instead of forcing every post to chase the same action.

How Instagram recommendations can expand beyond your followers

One of Instagram's more useful public explanations describes recommendation distribution as a staged process.

Eligible original Reels can be shown to a relatively small group of people Instagram expects may enjoy them, including non-followers. Better-performing content can then be exposed to progressively wider audiences.

This is important for smaller accounts because follower count doesn't need to be the entire starting audience.

It also explains why some Reels stop after a modest first wave of reach while others continue moving. There isn't necessarily a hidden penalty involved. The early viewer response may simply have been weaker than the alternatives competing for recommendation inventory.

Instagram doesn't publish a universal number for the size of these test audiences, and there is no trustworthy “get 50 likes in 30 minutes and Instagram unlocks the next tier” formula.

Don't build a strategy around invented thresholds.

Trial Reels give you a cleaner way to test non-follower response

Trial Reels are particularly interesting for creators trying to understand discovery because they can initially expose a Reel to non-followers.

Meta said in 2025 that, in an internal analysis of more than 400,000 creators using Trial Reels, 40% began posting Reels more often afterward. Among creators who increased their posting frequency, 80% saw an increase in non-follower Reel reach in the measured period.

That isn't proof that Trial Reels cause an 80% reach increase. Meta explicitly notes that performance isn't guaranteed, and the statistic describes a subgroup of creators who changed their posting behavior.

The feature is still useful as an experiment. If you have an idea that feels slightly outside your normal content, testing it against non-followers can give you information without making assumptions based entirely on your existing audience.

Why your Instagram reach may suddenly drop

A reach decline is easier to debug when you stop treating every drop as the same problem.

1. Check recommendation eligibility first. Open Account Status and look for restrictions affecting recommendations. If the account isn't eligible for discovery, solve that before changing your content schedule.

2. Split follower and non-follower performance. A decline concentrated among non-followers points toward a different problem than a decline across your established audience.

3. Compare like rates relative to reach. If reach fell but likes per reach stayed healthy, the people who received the post may still have liked it. Distribution and viewer response aren't the same metric.

4. Check sends. If your growth content gets likes from followers but almost nobody sends it, you may have built something your community approves of without giving Instagram much evidence that strangers will care.

5. For Reels, inspect retention rather than guessing. Look for the moment people stop watching. Then edit the next Reel around that specific failure point.

6. Audit originality. If your strategy depends heavily on downloaded clips, reposted memes, screenshots, compilations, or other creators' media, review whether you are adding substantial original value.

7. Compare several posts, not one. One weak Reel is noise. A repeated change across comparable posts is much more useful.

If likes themselves suddenly disappear or appear inconsistent, that can also be a display or account issue rather than a ranking change. Our guide to seeing Instagram likes covers the current visibility options, while Instagram Comments Not Showing explains common comment-side problems.

A better way to test the Instagram algorithm

You can't run a perfect laboratory experiment on Instagram. Audience composition changes, topics differ, competitors publish at the same time, and Instagram itself keeps updating its systems.

You can still make your tests much less noisy.

Start with a group of posts that solve the same kind of problem for the same audience. Keep the topic family reasonably consistent. Then change one important element at a time.

For example, test the opening of your next three Reels while keeping their approximate topic and length similar. On another set, keep the opening style stable and change the type of payoff: checklist, demonstration, result, or opinion.

Record at least:

• views;
• accounts reached;
• follower versus non-follower reach;
• average watch time for video;
• likes per reach;
• sends per reach;
• saves per reach when relevant;
• profile visits and follows when the content is supposed to create account growth.

Don't change the hook, topic, video length, editing style, CTA, and posting time at once and then call the winner an “algorithm test.” You won't know which difference mattered.

For ideas that match different account stages and content goals, see our Instagram Post Idea guide.

What I wouldn't build an Instagram strategy around in 2026

A magic posting time

Recency can matter, particularly where Instagram is choosing among fresh content, and publishing while your audience is active can help people encounter the post sooner. That doesn't create a universal best hour.

Your own audience data is more useful than a chart saying everyone should post at 7:14 PM on Tuesday.

Raw follower count

Follower count affects the size of the audience you've already built. It doesn't tell you how many people Instagram will recommend the next Reel to.

Our own public-account research is a good illustration of how widely engagement varies even before private reach metrics enter the picture.

Raw likes

Likes are useful. Likes without reach are difficult to interpret.

Use likes per reach for your own posts. When evaluating another public account where reach isn't available, compare multiple recent posts and account size rather than treating one visible total as a performance score. Our Instagram engagement rate guide explains the limitations in more detail.

Hashtag tricks

Hashtags can help describe and organize a topic, and Instagram Search is its own discovery system. That doesn't make a particular number of hashtags a substitute for viewer response.

I would spend more time making the topic obvious in the content itself than searching for a secret hashtag combination.

One “winning” format

Reels are useful for discovery because Instagram has built a large recommendation environment around them. Carousels can work extremely well for material people want to consume slide by slide, save, or reference. Stories are strong for repeated interaction with an existing audience.

Choose the format based on the job.

A practical Instagram algorithm checklist for 2026

Before publishing, ask whether someone who doesn't know the account can understand the post without additional context. Make the subject clear early. For video, give the viewer a reason to keep watching before spending time on an introduction.

After publishing, separate reach from response. Check who Instagram showed the post to, then inspect what those people did. Track watch time, likes per reach, and sends per reach over multiple comparable posts.

If discovery is the goal, pay special attention to non-follower reach and sends. If community retention is the goal, follower response and relationship-driven formats such as Stories may tell you more.

And if reach collapses unexpectedly, check Account Status and originality before rewriting your entire content strategy.

Frequently asked questions

What are the most important Instagram algorithm signals in 2026?

Instagram uses many signals and different ranking systems for different surfaces. For practical content analysis, Adam Mosseri has highlighted watch time, likes, and sends. Likes are slightly more important for connected reach, while sends are slightly more important for reaching people who don't follow you.

Does Instagram favor Reels in 2026?

Reels have a strong discovery function and are frequently shown to non-followers, so they can create substantial unconnected reach. That doesn't mean every Reel automatically receives more reach than every carousel or photo. Viewer response, recommendation eligibility, topic, audience fit, and competition still matter.

Are saves more important than likes?

There is no official universal rule saying one save is worth a particular number of likes or that saves are Instagram's strongest signal. Saves can be highly useful for content people want to revisit, while Instagram has specifically told creators to monitor watch time, likes per reach, and sends per reach.

Do shares increase Instagram reach?

They can be important, particularly sends through direct messages. Mosseri has specifically identified sends per reach as an important metric for unconnected reach. That doesn't guarantee wider distribution for every shared post.

Does follower count affect Instagram reach?

Follower count determines the size of your established audience, but Instagram can recommend eligible content to people who don't follow you. In LikeWave's August 2026 study of 137 public accounts, follower count had only a moderate positive rank correlation with absolute average likes, showing why account size alone is an incomplete performance benchmark.

Why is my Instagram reach suddenly down?

Check whether the decline affects followers, non-followers, or both. Then review Account Status, recommendation eligibility, originality, watch-time changes, likes per reach, sends per reach, and several comparable recent posts. A reach drop doesn't automatically mean the account has been “shadowbanned.”

Is original content more important in 2026?

Yes, particularly for recommendations. Meta said that 75% of Instagram recommendations in the United States were coming from original posts in Q4 2025. Instagram also expanded its approach to unoriginal aggregator content in 2026 to include photos and carousels.

Sources and research notes

This article combines Instagram and Meta's published ranking guidance with LikeWave's own public-account research. Meta's public documentation describes how its systems work at a high level; it does not publish every model weight or provide creators with a formula for predicting reach.

• Instagram: Instagram Ranking Explained
• Meta Transparency Center: Explaining Ranking
• Meta: How AI Influences What You See on Facebook and Instagram
• Meta: 2026 — AI Drives Performance
• Meta: Inspiring Creativity That Brings People Together
• Social Media Today: Instagram ranking signals discussed by Adam Mosseri
• TechCrunch: Instagram's April 2026 update on aggregator accounts
• LikeWave: Instagram Likes Statistics 2026 — 137 public accounts analyzed

LikeWave is not affiliated with Instagram or Meta. Ranking systems change over time, and individual account performance is never guaranteed.