How the Instagram Reels algorithm actually decides who sees your video

Nobody outside Meta has read the code. But between Instagram’s own public statements, the metrics they chose to expose, and a great deal of observable behaviour, the machine can be described accurately at the level that changes what you film.

Last updated 12 August 2026

Most explanations of the Reels algorithm are lists of things you should do, presented as if someone read the source code. Nobody outside Meta has. What we do have is a decade of Instagram's own public statements, the metrics they chose to expose in the app, and a very large amount of observable behaviour. That is enough to describe the machine accurately at the level that changes what you film.

The single most important thing: ranking is per-video

Instagram does not assign your account a score that determines how your posts do. Each Reel is evaluated on its own. This is why a small account can post one video that reaches a million people, and why an account with a hundred thousand followers can post something that reaches four thousand.

Everything downstream follows from this. Your follower count is not a reach multiplier. Your last flop does not tax your next post. The video is the unit.

How a Reel gets distributed, step by step

1. A seed audience sees it. Some of your followers, and a small group of non-followers whose behaviour suggests the subject fits them.
2. Their response is measured. Not their opinion — their behaviour. How long they watched, whether they finished, whether they replayed, whether they sent it to someone.
3. A bigger group sees it, or it stops. If the sample responds well, the Reel goes to a wider and less closely-matched audience. That audience is a harder test, which is why videos often plateau at a certain size.
4. The loop repeats until it fails. Every video eventually stops, when the next audience out is unrelated enough that the numbers fall.

A video that "goes viral" is one that kept clearing the bar as the audience got progressively less targeted. That is the whole trick, and it explains why niche-specific content often outruns broad content: it clears the early rounds so decisively that it can afford to degrade later.

The signals, in rough order of weight

Instagram has been fairly consistent publicly about which actions matter most, and what they say lines up with what accounts observe.

• Sends. Instagram has said plainly that sends per reach is the metric they would pick if they could only look at one. It is the hardest signal to game and the clearest proof a video is worth another person's attention.
• Watch time and completion. Both the raw seconds held and the proportion of the video finished. A short video finished is treated differently from a long video abandoned at the same timestamp.
• Replays. Rewatching is a strong quality signal and part of why loop-friendly edits do well.
• Saves. Intent to come back — strong for anything instructional.
• Likes and comments. Real but weaker than the above, and comments matter more when they are conversations rather than emoji.
• Follows from the post. Rare and heavily weighted when it happens.

The practical reading: optimise for the video being passed on, not for the video being liked. Those are different edits. A likeable video is agreeable. A sendable video is useful, funny enough to be a gift, or precise enough about a shared experience that forwarding it is a form of speech.

What matters less than people think

• Posting time. It affects how quickly the seed audience assembles, not whether the video travels. A good Reel posted at an awkward hour still finds its audience; a weak one posted at the optimal minute still doesn't.
• Hashtags. A categorisation hint. Three or four relevant ones are fine. Thirty is not a strategy.
• Follower count. See above — per-video ranking. It changes the size of the seed group, not the ceiling.
• Consistency streaks. Posting regularly gives you more attempts and more data. It does not earn you goodwill from a ranking system that has no memory of your diligence.

What actively suppresses reach

These are the ones Instagram has been explicit about, and they are worth knowing because they are cheap to avoid:

• Visible watermarks from other platforms. Recycled TikTok exports with the logo still on them are demoted.
• Low-resolution or heavily letterboxed video. Anything that reads as recompressed.
• Engagement bait. Explicitly asking for comments, shares or follows as the mechanism of the post.
• Reposted content you did not make, without meaningful addition.

How to use any of this

The honest conclusion is that the algorithm is not the lever. It is a fairly plain measurement device: it finds out whether people want to watch a thing and then acts on the answer. Every hour spent trying to outwit it would be better spent on the two seconds that decide whether the sample audience stays.

The one genuinely useful algorithmic insight is the per-video framing. It means the correct unit of study is a video, not an account. When you look at a creator who is beating you, do not ask what they do generally. Find the specific posts that beat their own average, and work out what those have in common that their normal posts don't. That difference is the thing worth copying, and it is usually structural — the shape of the opening, the length, where the payoff sits — rather than topical.

Questions people actually ask

How does the Instagram Reels algorithm work?

Instagram ranks each Reel individually rather than ranking your account. It shows a new Reel to a small seed audience, measures how they behave — watch time, completion, replays, sends and saves — and expands distribution to progressively wider and less targeted audiences for as long as the response holds up.

What is the most important Reels metric?

Sends per reach. Instagram has said publicly that if they could look at a single metric it would be this one, because forwarding a video to someone is the hardest signal to fake and the strongest evidence the content is worth another person's time.

Does follower count affect Reels reach?

Only in that it changes the size of the initial seed audience. Because ranking is per-video, a small account can reach far beyond its follower count, and a large account regularly posts Reels that reach a fraction of its followers.

Does posting time matter for Reels?

Much less than people assume. Posting time affects how quickly the seed audience assembles, not whether a Reel travels. A strong Reel posted at an awkward hour still finds its audience.

What gets a Reel's reach suppressed?

Visible watermarks from other platforms, low-resolution or recompressed video, explicit engagement bait, and reposting content you did not create without meaningful addition. Ordinary content getting low views is a performance result, not a suppression.

Do hashtags help Reels reach?

They act as a categorisation hint, so a few relevant tags are worth adding. They do not rescue a video people swipe past, and large hashtag blocks have no additional effect.