Skip to content

What Instagram says ranks Reels: watch time, likes and sends

What Instagram's own head has said about how Reels are ranked, what that means for the scripts you write, and how to use trial reels to test ideas on new viewers.

By Ebenezer Don4 min read

There is a lot of guesswork written about the Instagram algorithm. There is also a fair amount of what Instagram itself has said, mostly through Adam Mosseri, who runs the app, in videos and posts aimed at creators. This piece sticks to the second kind, and then turns it into things you can do in the script.

Instagram changes often, so treat everything here as "as of late 2026" and check Instagram's own creator resources for anything newer.

The three signals Instagram named

In January 2025, Mosseri said the three most important signals for ranking are watch time, likes and sends (shares by direct message). He suggested creators look at three numbers in their insights to understand performance:

  • Average watch time
  • Likes per reach
  • Sends per reach

He also said the weights differ slightly by audience. Likes count a little more for people who already follow you. Sends count a little more for people who do not. That second point matters most for growth: a Reel that people send to a friend is the kind most likely to be shown to strangers.

Note what is in those numbers: they are rates, not totals. A Reel that reached 2,000 people and got 100 sends is doing better, by this measure, than one that reached 50,000 and got 300.

What each signal asks of the script

Watch time: the first second and the middle

Watch time starts with the hook. If the first spoken words and the first frame do not make the idea clear, most people are gone before the second sentence. The structure in how to script a TikTok, Reel or Short is built around this: hook, turn, payoff, and stop.

The middle is the other half. Every line should raise the next question. Lists lose people; each point leading to the next keeps them.

Likes: give something worth rewarding

A like is a small "that was good". People tend to give it after a payoff: a result, a laugh, a useful tip that worked. If the payoff is buried at the end of a long sign-off, many viewers never see it.

Sends: write for the friend who needs this

Sends are the most useful signal to write for, because they reflect a specific thought: "this is for them". Videos get sent when they:

  • Solve a problem someone you know has. "Send this to the person who always burns the rice."
  • Name a shared experience precisely. The more specific, the more people think of one friend.
  • Settle an argument. Comparisons, myths, "you have been doing this wrong".

You cannot ask for sends into existence, but you can make the reason to send obvious. If you can picture the exact message someone would type when sending your Reel, the idea is probably a good one.

Originality matters

Instagram has taken several steps to favour original content over reposts. In 2024 it said aggregator accounts that repeatedly repost other people's content would be removed from recommendations, and reposts would carry a label linking to the original creator.

For creators, the practical meaning is simple: your own footage of your own idea is what Instagram wants to recommend. A Reel that reuses someone else's clip with text on top is working against the system, not with it.

Trial reels: test on people who do not follow you

Since December 2024, professional accounts can post trial reels. A trial reel is shown first to people who do not follow you, and stays off your grid and your followers' feeds. After about a day you get its numbers, and you can then share it with your followers, or set it to share automatically if it does well.

This is the best testing tool Instagram has given creators, and it changes how you can use scripts:

  1. Write two hooks for the same idea. Record both.
  2. Post one as a trial, or both as separate trials.
  3. Compare watch time and sends per reach on non-followers, the audience that matters for growth.
  4. Share the winner with your followers, and keep the losing hook as a lesson.

Instagram has said trial content can take longer to gain traction, since followers do not see it first. Give it a day before reading the results.

Length: up to 3 minutes, but not by default

Reels can run up to 3 minutes since January 2025, up from 90 seconds. When Instagram announced the change, it had previously told creators that shorter Reels tended to perform better. Both can be true: some ideas need two minutes, and a 25-second idea stretched to two minutes loses watch time.

Write the idea first, then let the length follow. If a Reel's retention graph drops sharply at one point, that is where the idea ran out.

A weekly habit that uses all of this

  1. For every Reel, note average watch time, likes per reach and sends per reach.
  2. Sort by sends per reach. Those are the ideas most likely to reach new people. Make more like them.
  3. Use one trial reel a week to test a new hook or a new format on non-followers.
  4. Rewrite your weakest opening from the week and post it again later as a trial.

None of this needs a trick. It needs clear ideas, openings that land in a second, a payoff worth liking, and a reason for someone to send it to a friend.

CopyMe, the AI studio for talking videos

  • Shorts, Reels and TikTok

    How to script a TikTok, Reel or Short that people finish

  • Creator news

    Instagram's Edits video editor arrives on Mac: creator news, 6 Nov

  • Creator news

    Instagram's originality rules: what counts as original in 2026