Instagram Algorithm 2026: Read Your Reels Like Edits' New AI

Instagram shipped an AI that explains why your Reels work, but only in the US. Here is how to run the same analysis from Latin America.

What changed this week on Instagram and YouTube

On September 30, Instagram launched an AI assistant inside Edits, its video editing app. It reads your account metrics (follows, views, retention, likes, shares), cross-checks them against your comments and what is trending, and tells you why one Reel worked and another did not.

Brett Westervelt, who leads Edits, put it plainly: creators want a tool that handles the analysis, not one that does the creative work for them. It is free with a daily prompt limit, and the Meta One subscription unlocks more usage.

One detail matters if you live in Mexico, Colombia or Argentina: for now, it is only available in the United States. Meta says it will reach more markets "in the coming weeks," but there is no date for Latin America or Spain.

A day later, on October 1, YouTube announced that its recommendation system will prioritize original Shorts and cut reach for channels that re-upload other people's clips without adding anything. Two platforms, one week, one message: the algorithm rewards your own content that people watch to the end and share.

The good news: you do not have to wait for the AI to reach your country. The signals that AI analyzes are public, and you can compute them with a spreadsheet and 20 minutes a week.

The three signals Instagram has actually confirmed

Adam Mosseri, head of Instagram, has repeated throughout 2026 that a Reel's reach depends on three main signals:

• Watch time: seconds viewed, including replays and completion rate.
• Likes per reach: likes divided by accounts reached, not the raw count.
• Sends per reach: how many people forwarded your Reel by DM, again as a share of reach.

The key phrase is "per reach." Instagram measures rates, not totals. That is why an account with 3,000 followers in Lima can beat one with 300,000 if its content gets shared more, proportionally.

Sends weigh more than likes. When someone forwards your Reel to a friend, they hand Instagram a direct recommendation to a person who does not follow you. It is the shortest path to new people.

Do the analysis yourself (no AI required)

This is exactly what the Edits assistant does: it takes your numbers and looks for patterns. You can do the same with Instagram Insights and a table. Example: a personal-finance coach in Bogotá who posted three Reels in the same week:

Looking only at likes, Reel B looks like the winner at 7.5% likes per reach. But almost nobody forwarded it, and half the viewers dropped off before the midpoint. Reels A and C were shared 7 to 8 times more, proportionally, which is why they reached 3 to 4 times more people.

The pattern is clear: specific content, with a number or a concrete mistake in the hook, gets sent by DM. Lifestyle content gets likes but does not travel. That is the "insight" the AI would hand you; you just pulled it in five minutes.

How to build your table every week

1. Open Insights for each Reel and copy reach, likes, sends and average retention.
2. Divide likes and sends by reach. Write down the percentage.
3. Flag the two Reels with the highest sends/reach and write one line about what they have in common.
4. Make two new Reels with that pattern. Repeat next week.

If your Reels generate DMs like "where do I buy your course?" and nobody answers within the first hours, that interest cools off. Every send that turned into a conversation and got no follow-up is a lost lead the algorithm had already handed you.

Originality: the rule Instagram, Facebook and YouTube now share

YouTube's October 1 announcement is not an outlier. Instagram limited reach for reposted content in April 2026 and Facebook did so in July 2025. All three platforms are converging on the same standard.

Rene Ritchie, YouTube's creator liaison, was explicit about what does not count as original:

• Voice-over narration describing what is already on screen.
• Minor technical edits (trims, filters, speed changes).
• Template-based bulk changes.

And what does count: your own commentary, analysis or storytelling. Cutting your own long-form videos into Shorts remains fine. YouTube clarified the change affects distribution, not monetization, and that channels shifting to original content will have their recommendations re-evaluated.

For a Latino creator, this is a competitive edge. If you teach in Spanish with your own examples, your accent and your local context, nobody can re-upload that as theirs. "Generic translated" content is exactly what loses reach now.

Test before you post: Trial Reels

Instagram has a feature built to solve the fear of "burning" your audience with an experiment: Trial Reels. They show only to people who do not follow you; if they work, you push them to your profile and followers.

The numbers Instagram reported are worth knowing:

• 40% of creators who used Trial Reels ended up posting more often.
• The feature drove an 80% increase in Reels reach among non-followers.

Since April 2026 you can schedule them. That matters if your audience is split between Mexico City, Madrid and Buenos Aires: you can launch the test at each zone's peak hour and compare sends/reach for two different hooks on the same topic.

Use them like this: same topic, two different opening seconds. One with a question ("Do you know how much you lose paying installments with interest?") and one with a fact ("Paying in 12 installments can cost you 30% more"). After 24 hours, post the one with the better sends/reach to your followers.

One caveat: a Trial Reel reaches strangers, so judge it by sends and retention, not by comments from people who already know you. If two variants tie, keep the one that holds attention past the midpoint; that is the number Instagram weighs first.

From reach to a list: move your audience to a channel you control

No algorithm belongs to you. Two platforms reminded everyone of that within 48 hours this week. A Reel's reach is useful if part of that audience lands somewhere you decide when and how to talk to them: WhatsApp, email or your own community.

Back to the Bogotá coach, with 91,000 accounts reached in a week:

That is $1,276 USD in a month from a single workshop, with no ad spend. And the list of 546 contacts is still yours for the next launch.

What changes when you sell in Latin America

• Price in local currency: the same $29 USD workshop can show as MXN, COP or PEN at checkout. The buyer understands the price without converting.
• Local payment methods: in Mexico many people pay at OXXO; in Colombia, with PSE or Nequi. If you only take international cards, some of those 44 buyers never finish paying.
• Installments: a $149 USD course in 3 or 6 installments converts differently from a single payment, especially in Argentina and Colombia. Ideally, those installments should not come out of your pocket.
• Getting paid in USD: if you live in a high-inflation country, receiving dollars protects what all those Reels earned you.

Your 7-day plan

1. Monday: build the table with your last 10 Reels. Compute likes/reach and sends/reach.
2. Tuesday: find the pattern in your top 3 by sends/reach. Write 5 new hooks with that pattern.
3. Wednesday: record 2 original Reels (your voice, your example, your country). No templates, no borrowed clips.
4. Thursday: post one as a Trial Reel with two hook variants, scheduled for your zone's peak hour.
5. Friday: post the winner to your followers. Answer every DM in under an hour.
6. Saturday: message your list (WhatsApp or email) with the Reel and a clear offer in local currency.
7. Sunday: update the table. Compare this week's sends/reach against last week's.

When the Edits assistant reaches your country, you will use it with an advantage: you will already know how to read your numbers. Until then, the spreadsheet does the job.