How the Instagram algorithm works (no smoke)

“The algorithm buried me”, “the algorithm changed”, “you have to hack the algorithm”. It gets said a lot and very little is known, and in between there's an entire industry selling courses to decode it. Here's the opposite: only what Meta has said publicly, the myths that cost the most, and the part that actually depends on you.

Nobody outside knows the algorithm (and whoever says they do is selling you something)

The system that decides what Instagram shows each person is internal to Meta, changes all the time and isn't published anywhere. The only reliable thing is what Instagram itself explained on its blog and in its team's videos — which is less than you'd like, but more than most people use. Everything else, the PDF with “the 2026 algorithm rules” or the course that promises to decode it, is guesswork with nice design. And notice the detail: whoever claims to know the algorithm is always about to charge you for something.

This isn't bad news. It means your competition — the deli around the corner, the nutritionist with forty thousand followers — doesn't know it either. Nobody plays with the full map: you play by reading signals, and the signals are public.

There isn't ONE algorithm: feed, Reels, stories and Explore rank differently

Instagram said it in so many words: each section of the app has its own ranking system, because people open each one looking for something different. Stories prioritize relationship: they show you first the accounts you already interact with. Reels and Explore do the opposite — they're built to show you accounts you don't follow. The feed mixes: people you know plus suggestions.

The practical consequence is huge: your Reel competes for new people and your story works the clientele you already have, so judging them with the same yardstick makes no sense. Which format fits which goal we already took apart in “Reels, carousels or stories: which one fits your goal”. Keep this: you don't post on Instagram, you post on four different playing fields.

The signals Meta did name: retention, interaction, saves, relationship

When Instagram explained how it orders content, it named concrete signals. No weights and no percentages — it never published those, and any number you see around is made up. The ones that come up most in its explanations are four:

• Retention: how long people stay watching your piece. It's the signal Instagram's own team talks about most when explaining Reels.
• Interaction: likes and above all comments. A video that sparks conversation tells the system something is happening there.
• Saves and shares: someone saving your price carousel or sending your Reel to a friend is saying “this is worth it” more strongly than a like.
• Prior relationship: if that person already interacted with your account, they're more likely to see the next thing you post. Replying to comments isn't just good manners: it's a signal.

Your first 3 seconds weigh more than any hashtag

Look where most people put their energy: twenty minutes choosing hashtags, two thinking about how the video starts. It's exactly the reverse proportion. Hashtags help Instagram understand what your piece is about — they're a label, not a push. Retention, on the other hand, starts being decided in the first second: if people leave before the third, nobody's left to generate the other signals. Without retention there are no saves, no shares, nothing.

That's why the hook is the biggest lever you have over the algorithm without touching the algorithm. On how to build one —formulas, mistakes with their fix, the three tests before shooting— there's an entire guide on this blog: “How to write a good hook for your videos”. If reach keeps you up at night, start there.

The myths that cost you money and time

The four we hear most, with what's actually known about each one next to it:

• The hashtag shadowban. Instagram has publicly denied it more than once. What does exist is content that breaks the recommendation guidelines and gets shown less — and you can check that yourself in the app, under “Account status”. Almost always the explanation for the dip is less conspiratorial: the piece didn't retain.
• Magic posting times. There's no universal hour published by anyone. What exists is YOUR time: the moment your audience has the phone in hand. A store selling to moms doesn't post at the same time as a bar. Test and look at your numbers.
• “Ads switch off your organic reach”. There's no official source for that. What does happen is that viral gets confused with paid: many “success stories” you're shown have money behind them. To learn to tell them apart you have “Is it viral… or is it paid?” on the site.
• “If you don't post every day, it punishes you”. Nobody at Meta said that. Posting less doesn't penalize you: there are simply fewer pieces of yours competing. Frequency is math and stamina, not punishment.

What to look at in your numbers: the curve, not the counter

You can't see the algorithm, but you can see exactly what the algorithm sees of your pieces: retention, saves, shares. That's your only source of truth, and it's worth more than any course. The fine reading of the curve —where people drop off and what each drop means— we already covered in “How to get your first 1,000 followers”, so here goes only the rule that follows from this article: compare each piece against its own playing field. A Reel is measured against your other Reels, not against a carousel, because as we saw, the same system doesn't rank them.

And prioritize strong signals over vanity ones: a piece with few likes but many saves is flagging a topic to repeat. The follower counter is the last number to move; the signals arrive weeks earlier.

The only hackable thing: do more of what already worked for YOUR audience

After discarding the shortcuts there's a single strategy left that doesn't depend on guessing anything: look at which of your pieces generated retention, saves and conversation, understand why, and do more of that. The algorithm ends up being a mirror of your audience — “the algorithm” doesn't like you, the people it measures do.

For that part Gancho gives you hands. The Reels & TikToks Analyzer diagnoses a piece: you paste the script or the caption and it tells you what's weak in the hook, the structure and the copy — it doesn't play the video or invent metrics it can't see. The Profile Analyzer gives you a reading of your account or a reference's: bio, tone, content line. And when you know what to repeat, the Hook Generator gives you 12 hooks so you don't start from scratch. What the algorithm rewards is built piece by piece — and publishing is still yours.