Instagram Growth

How the Instagram Reels Algorithm Works in 2026

What Meta and Mosseri have confirmed about Reels ranking in 2026, what is folklore, and how to read your own numbers.

By Piyush Sachdeva · Oct 16, 2026 · 7 min read

How the Instagram Reels Algorithm Works in 2026

What changed this year is the gatekeeping, not the signals. Instagram still ranks Reels on whether people watch, send and like them, and Adam Mosseri's three numbers to track are average watch time, sends per reach and likes per reach. In 2026, originality rules and viewer topic controls decide who gets tested at all.

Updated: October 2026

TLDR

Reels is ranked on its own, mostly for viewers who do not follow you, so a new Reel is first tested on a small cold audience.
Mosseri's three signals to watch are average watch time, likes per reach and sends per reach. Sends matter slightly more for strangers.
The popular claim that sends count 3 to 5 times more than likes has no named source.
Reposts, watermarks and mostly unoriginal accounts lose recommendations. Original work is the entry ticket.
Read your own ratios by comparing your Reels with each other, not with a viral benchmark.

Quick answers

One Reels algorithm? No. Meta says each part of the app, Reels included, has its own ranking system.
Signals Mosseri named? Average watch time, likes per reach and sends per reach (January 2025).
Hashtags and Reels ranking? Meta's Reels ranking explainer lists no hashtag signal.
Watermarked Reels? Instagram says it tries to make them less visible.
Reels shadowban? Meta says limiting reach without explanation is not its intent, and Account Status shows recommendation eligibility.

How does Instagram decide which Reels to show?
Which signals matter most for Reels?
Do sends really count 3 to 5 times more than likes?
What does Instagram demote or refuse to recommend?
What changed in 2025 and 2026?
How do you read your own Reel numbers?
Where does a scheduler fit?
Frequently asked questions
Sources

How does Instagram decide which Reels to show?

Meta's 2023 ranking explainer says Reels works a lot like Explore. Most of what a viewer sees comes from accounts they do not follow, so Instagram first sources Reels the viewer might like, then orders them by how interesting it thinks each one is to that person. To judge "interesting", Meta says it surveys viewers on whether a Reel was worth their time.

The most important predictions it lists are whether the viewer will reshare the Reel, watch it all the way through, like it, or tap through to the audio page. That last one is a proxy for being inspired to make your own.

The test comes next. When Instagram announced its originality changes in April 2024, it described eligible Reels going to a small audience whether or not they follow you, then to wider groups as the best performers prove themselves (reported here). Follower count is not the gate. The first small audience is.

Before any of that, a Reel has to be recommendable. Content that breaks Meta's Recommendation Guidelines can still reach your followers, but Instagram tries not to show it to strangers. Account Status in your settings shows whether your account and content are eligible. If reach to non-followers looks stuck, check there first, and read our shadowban guide for the symptoms that get misread.

For the other surfaces (Feed, Stories, Explore), our broader Instagram algorithm guide covers the whole system. This post stays on Reels.

Which signals matter most for Reels?

On January 22, 2025, Mosseri named the three signals that matter most for ranking: watch time, likes and sends, as Social Media Today reported. He told creators to look at average watch time, likes per reach and sends per reach in their insights. The "per reach" part matters. These are ratios, so a Reel seen by 500 people and one seen by 50,000 can be compared fairly.

He added that likes count slightly more for people who already follow you, and sends are "slightly more important for unconnected content". Comments were not on his list. Meta's 2023 explainer does count a viewer's own commenting as part of their activity history, so comments are not ignored. They just were not named in the top three.

Here is how much of the Reels advice you read traces back to something Meta or Mosseri said.

Claim	Where it comes from	Our read
Reels has its own ranking, mostly for non-followers	Instagram ranking explainer, May 2023	Official
Reshare, finish, like and audio tap are the key predictions	Same explainer	Official, three years old
Average watch time, likes per reach and sends per reach are the top three	Mosseri video, January 22, 2025	On the record
Sends count 3 to 5 times more than likes	Tool blogs, no named source	Unsourced
Two-way conversation became a ranking signal in March 2026	Tool blogs only	Not found in any Instagram source
Unoriginal accounts lose recommendations	Instagram creator guidelines, April 2026 reporting	Official

Do sends really count 3 to 5 times more than likes?

You will see "sends count three to five times more than likes" in a lot of 2026 guides. One analysis that went looking for the origin found that nobody names a source. Mosseri's own January 2025 wording was "slightly", not a multiple.

Our read: sends are a strong signal for reaching strangers, which fits Meta's own list, where reshare comes first. The multiplier looks like a number somebody made up and everybody repeated. Nobody outside Instagram knows the weights, and they probably shift by viewer and by surface. Aim for a Reel someone would forward to a friend, and stop doing arithmetic on it.

The 2023 explainer lists Reels Instagram tries to make less visible: low resolution, watermarked, muted or bordered, mostly text, or already posted on Instagram. The fix is dull but real. Export without another app's watermark, in full frame, with sound or captions.

Originality is the bigger gate. Instagram's original content guidelines say unoriginal content may be ineligible for recommendation to non-followers. An account that mostly posts unoriginal Reels, photos or carousels can lose recommendations until most of its recent posts count as original again over a rolling 30-day period. Borders, watermarks, speed changes and credit-only reposts are listed as immaterial edits. Your followers still see your posts either way.

Press coverage of the 2024 announcement put the line at roughly ten unoriginal posts in 30 days (ZDNet). Instagram's current page gives no number, so we would not lean on that figure. On April 30, 2026, Instagram applied the same logic to photos and carousels, per Tubefilter.

What changed in 2025 and 2026?

January 2025: Mosseri names the three signals to track (see above).
Trial Reels, rolled out from December 2024: a Trial Reel is hidden from your followers and shown to non-followers, metrics appear after about 24 hours, and it can auto-share if it hits a threshold within 72 hours, per Instagram for Creators.
December 10, 2025: Instagram launches Your Algorithm, which shows viewers the topics shaping their Reels and lets them adjust them. It started in the US, per TechCrunch. Tool blogs report later expansion to Explore and Feed, which we could not confirm in an Instagram source.
April 30, 2026: the originality penalty reaches photos and carousels.

Our read on Your Algorithm: when viewers pick topics, a clear niche is easier for Instagram to match with the right people. Nothing official says topic choice outweighs the signals above, so it is a reason to stay consistent, not a new trick.

How do you read your own Reel numbers?

Take two made-up Reels, both 20 seconds long and each reached by 8,000 accounts. The numbers are invented to show the arithmetic. They are not benchmarks.

Reel	Avg watch time	Likes	Sends	Likes per reach	Sends per reach
A	6 s	160	80	2.0%	1.0%
B	11 s	120	240	1.5%	3.0%

Reel A wins on likes. Reel B was watched almost twice as long and sent three times as often per viewer reached. B is the one to remake. For scale, Metricool's 2026 study of about 24 million posts put the average Reels watch time at 8.5 seconds.

What you see	Likely cause	Try this
Watch time below your usual	Slow opening	Cut the warm-up and test the first line with the hook timing analyzer
Likes fine, sends low	Relatable but not forwardable	Make it useful or funny enough to send to a specific person
Reach mostly followers	Not recommended to strangers	Check Account Status, and split Accounts Reached by followers and non-followers in Insights

For overall engagement rate, use our free engagement rate calculator. Compare Reels with your own last ten before you compare them with anyone else's.

Where does a scheduler fit?

We found nothing in Meta's Reels ranking material that lists the publishing tool as a signal. Reach comes from the inputs above. ReelDrop is our product: it schedules Reels through the official Instagram API, lets you pick the cover frame, and supports Trial Reels (3 a month on the free plan, unlimited on Influencer and Creator). Later, Publer and Metricool also schedule Trial Reels, so that is not a ReelDrop exclusive. See the scheduler page or the schedule your first Reel doc.

Frequently asked questions

Does Instagram favor Reels over photos?

Reels are built for discovery, so most of what viewers see in Reels comes from accounts they do not follow. That gives a Reel a bigger pool of strangers to be tested on. It does not mean photos or carousels are penalized. Each part of the app has its own ranking system.

How long does Instagram test a new Reel?

Instagram does not publish a window for normal Reels. Trial Reels give the only official reference: metrics appear after about a day, and a Trial Reel can auto-share if it hits a performance threshold within 72 hours. Treat that as a Trial Reels detail, not a rule for every Reel.

Do hashtags affect Reels ranking?

Meta's Reels ranking explainer lists no hashtag signal. Instagram now limits captions to five hashtags and says fewer, more targeted tags work better than many generic ones. We cover the evidence in our hashtag count guide. Treat hashtags as context for what the Reel is about, not as a reach lever.

Is there a best length for a Reel?

Neither Meta nor Mosseri gives an ideal length. Instagram's creator page says Reels must be three minutes or less to be recommended, but that page is dated May 2023. Because watch time is judged per viewer, a Reel people finish or replay beats a longer one they abandon.

Will a TikTok watermark hurt my Reel?

Instagram says it tries to make watermarked Reels, and Reels already posted on Instagram, less visible in recommendations. Export your own footage without another app's watermark. Content you made yourself counts as original under Instagram's guidelines. The rule targets reposted work you did not create or materially change.

Does the Your Algorithm tool change what creators should do?

Your Algorithm lets viewers see and adjust the topics behind their Reels recommendations, which Instagram says come from their activity. For creators, a consistent topic is easier to match to interested viewers. Instagram has not said how topic choices weigh against watch time, likes and sends.

Sources

Instagram Growth Guides