INSTAGRAM RANKING · 2026 SERIES

9 primary sources

Instagram Ranking in 2026: What Mosseri's 3-Part Series Actually Says

Instagram does not rank every surface with one algorithm. Mosseri's September 2026 series describes separate systems that predict what each viewer is likely to value, while current Meta system cards document different models for connected Feed posts, recommendations, and Reels. For creators, the useful move is not to chase a secret weight. Make one clear content hypothesis, change one input at a time, and compare watch behavior, shares, and audience response over a fixed window. Viewers can also reshape recommendations with Following, Favorites, Not Interested, and other controls.

Written and researched by Kabo Editorial TeamReviewed by Emma HsiaUpdated October 30, 202610 min read

01 · THREE EPISODES, ONE READ

What Mosseri's three-part ranking series actually says

Adam Mosseri's three posts form a useful sequence, but they do not publish an Instagram ranking recipe. Part one sets up why ranking exists. Part two widens the discussion beyond engagement. Part three turns toward the controls people can use to shape what they see.

Mosseri's September 2026 Instagram ranking series. Dates follow Instagram's public post metadata.
Published	Focus	What to rely on
September 3, 2026	Why ranking exists	Instagram ranks different surfaces and makes predictions about what each person is likely to value. The episode is orientation, not a formula.
September 8, 2026	What changes the systems	The caption names engagement, originality, recency, breaking content, and larger AI models. It does not publish weights or a universal order.
September 17, 2026	Controls people can use	The third installment shifts to ways people can shape their experience. Its public page does not include a reliable transcript of every example.

Part two's caption names originality, recency, breaking content, and larger AI models alongside engagement. Those are categories to understand, not a ranked list. No episode gives a multiplier, a universal priority order, or an action that guarantees reach.

Instagram: Ranking series, part one Instagram: Ranking series, part two Instagram: Ranking series, part three

October 28: what should a new creator track?

In an October 28, 2026 interview clip with Colin and Samir, Mosseri describes originality, giving new content a chance to break out, and recency as considerations beyond engagement. For reaching people who do not follow you, he places particular emphasis on sends per reach. His advice to new creators lists these metrics in this order:

Sends per reach
Likes per reach
Watch time or average watch time

Treat that as advice about what to monitor, rather than a published formula or a guarantee that improving one number will increase distribution. Here, sends per reach means how often a post is sent by direct message relative to the accounts it reached.

Constructed example: If your own analytics reports 150 direct-message sends and 10,000 accounts reached, the ratio is 150 ÷ 10,000 = 0.015, or 1.5 sends per 100 accounts reached. That is a calculation, not a target or a percentage of unique people who sent the post. If your interface only shows total Shares, keep that label instead of assuming every share was a DM. Public views and follower counts cannot fill in missing reach or sends.

Colin and Samir with Adam Mosseri: October 2026 ranking interview

02 · SEPARATE SYSTEMS

How Instagram ranking works without one master algorithm

Instagram's 2023 explainer says the app does not use one algorithm for everything. Feed, Stories, Explore, Reels, Search, and other surfaces are used differently, so Instagram ranks each one with systems suited to that context. It is useful background, not a live changelog.

Instagram: Instagram Ranking Explained

Meta's current system cards make the separation more concrete. Connected Feed content, Feed recommendations, and Reels chaining have their own documented systems. Each follows a broad pattern: gather eligible candidates, read signals, predict possible viewer actions, then rank the candidates. The models and inputs can change over time.

Three documented systems, with examples rather than a full formula
System	What it ranks	Example predictions
Connected Feed	Feed posts from accounts a person follows	Enjoyment, time spent, skipping, profile time, and sharing by direct message
Feed Recommendations	Suggested Feed posts from outside that set	Predictions built from the person, post, account, and interaction context
Reels Chaining	The next Reels in a viewing session	Short skips, relative watch depth, reshares, off-platform shares, comments, follows, and audio reuse

Meta system card: Instagram Feed Meta system card: Instagram Feed Recommendations Meta system card: Instagram Reels Chaining

These examples explain what kinds of predictions exist. They do not reveal every input, the weight of any input, or why a particular post appeared for a particular person.

03 · PUT CONTROL IN THE RIGHT COLUMN

The control map: what creators, viewers, and Instagram decide

Ranking advice gets misleading when it mixes creator choices, viewer choices, and Instagram's decisions. Put a proposed tactic in the correct column before you act on it.

Constructed control map for Instagram ranking decisions
Owner	What this owner can change	What stays outside this owner's control
Creator	Topic, framing, original contribution, opening, pacing, format, caption, timing, and whether the post is eligible to be recommended.	A guaranteed audience, a fixed signal weight, another account's behavior, or how every viewer will respond.
Viewer	Following and Favorites choices, Not Interested, hiding, muting or unfollowing, and suggested-content settings where available.	The full candidate pool, model design, or the exact interpretation of every action.
Instagram	Candidate selection, signal collection, predictions, ranking, integrity checks, model updates, and product experiments.	Whether a creator's idea is useful to a particular person or whether that person chooses to watch, share, or follow.

Following and Favorites help a person choose more directly from accounts they care about. Not Interested, hiding, muting, unfollowing, and suggested-content settings provide other ways to shape the experience, with availability varying by product and account. These are viewer controls. They are not creator hacks.

Instagram: Instagram Ranking Explained Meta system card: Instagram Feed Recommendations

04 · ORIGINAL DOES NOT MEAN UNFAMILIAR

What originality means in 2026

Instagram's April 2026 update expanded originality protections from Reels to photos and carousels. Its distinction is about contribution and ownership, not whether a format has ever appeared before. A creator can use a familiar format while supplying their own material, analysis, demonstration, story, or point of view.

Meaningful transformation and a genuinely new perspective may count as original contribution. Small surface edits do not do that work by themselves. Instagram specifically says that adding a border, watermark, subtitles, or credit alone is not a meaningful enhancement to someone else's post.

Instagram for Creators: Rewarding Original Creators

A practical originality check before publishing
Question	Stronger evidence of contribution	Weak evidence by itself
What did you add?	Original footage, reporting, analysis, demonstration, story, or a materially transformed presentation	Border, watermark, subtitles, or credit only
What can the viewer learn here?	A distinct claim, lesson, example, or point of view	The same post with cosmetic packaging changes
Can you prove the rights?	Your source files, permission, or licensed material	An assumption that public means free to republish

05 · RUN A CLEANER TEST

A seven-day test plan that avoids fake certainty

Use the series as a reason to test your own controllable inputs, not as permission to reverse-engineer a hidden score. A useful experiment sets the hypothesis and baseline before publishing on Day 1, leaves the comparison untouched through Day 6, and applies the decision rule at the agreed Day 7 cutoff.

01
Set the rule, choose the baseline, and publish
Write one falsifiable hypothesis, choose the closest recent Reel with saved same-age metrics, and change one meaningful input. Define the Day 7 cutoff and decision rule before publishing, check account status, then record the publication time.
02–06
Protect the observation window
Do not stack edits or declare a winner early. Keep a short daily note of account notices, distribution context, and the Insights available to you, but treat every number as provisional until the agreed cutoff.
07
Compare at the promised age
At the Day 7 cutoff, compare both posts at the same age. Read watch behavior, shares, follows, and audience mix together, write at least two plausible explanations, then apply the decision rule: keep, reject, or retest the change.

SEVEN-DAY EXPERIMENT CARDCopy before you publish

Hypothesis:
Closest baseline with saved same-age metrics:
One changed input:
What stays constant:
Eligibility or account-status check:
Day 7 cutoff:
Primary viewer behavior to compare:
Decision rule:
Two competing explanations:
Keep, reject, or retest:

This will not create a laboratory-controlled result. It will create a better decision than changing five things and crediting whichever explanation is fashionable. If two similar uploads diverge, use the matched-age diagnosis for why the same video can get different views before declaring a penalty or algorithm change.

06 · KEEP THE TOOL BOUNDARY CLEAR

What public tools can and cannot tell you

A public Reel contains evidence about the creative itself. The Kabo Instagram Reel Analyzer can inspect visible frames, spoken audio, pacing, and timestamps. That is useful when your test changes the opening, proof order, or dead space.

It cannot access private Instagram Insights, retention, shares, recommendation eligibility, or the per-viewer predictions Meta used. It cannot tell you why Instagram ranked a specific Reel for a specific person. Pair creative analysis with your own account notices and Insights, then describe the result as an observation rather than an algorithm diagnosis.

07 · FAQ

Common questions about the Instagram algorithm in 2026

What is the most important Instagram ranking signal?

Instagram has not published one universal winner. Different surfaces use different systems and multiple predictions. Define the surface and viewer behavior before discussing a signal.

Are shares more important than likes?

Meta documents sharing and other actions as prediction examples, but it does not publish a universal multiplier. Compare shares with the rest of the post's context and your goal instead of treating one ratio as a ranking law.

Does copying a proven format hurt reach?

A familiar format is not the same as reposting someone else's work. Use your own material or add meaningful transformation and a distinct perspective. Cosmetic changes alone do not establish originality.

Can I tell why one Reel was recommended?

Not from public data. You can inspect the Reel, account notices, and your own Insights, then test plausible explanations. The exact per-viewer ranking decision remains Instagram's.

08 · CHECK THE SOURCE

What Instagram documents, and what it leaves out

Mosseri's three posts explain what the series covers. Instagram's ranking explainer and Meta's system cards show why different surfaces do not behave as one algorithm, while Instagram's creator guidance sets the originality boundary. The control map and seven-day test are Kabo frameworks for turning that guidance into decisions, not Instagram tools or formulas.

Meta's system cards show important examples, not every input or weight. Instagram's broader ranking explainer dates to 2023. These sources were last checked on October 8, 2026; recheck them after a product or policy change. The October interview was checked on October 29, 2026.