# Instagram algorithm: how it decides what each person sees

> What Instagram predicts in each part of the app, according to its own documentation, and how to diagnose a reel that is not reaching non-followers.


**Instagram ranks each part of the app (Feed, Stories, Reels and Explore) with separate systems that estimate what each person will do with each post: whether they will stop on it, share it, follow the account or scroll past.** Those estimates decide what shows up first. In 2023, Instagram explained that it works with [a variety of algorithms, classifiers and processes, each with its own purpose](https://about.instagram.com/blog/announcements/instagram-ranking-explained).

That is why "the algorithm cut my reach" is rarely a useful diagnosis. The question that helps is narrower: which part of the app stopped showing your content, and to whom?

## What each part of the app tries to predict

Meta publishes system cards that describe what each ranking system predicts. The cards for [Feed](https://transparency.meta.com/features/explaining-ranking/ig-feed/) (updated June 2026) and [Reels](https://transparency.meta.com/features/explaining-ranking/ig-reels-chaining/) (November 2025) are the most recent; Stories and Explore follow the 2023 explanation.

| Surface | Who it shows | What it tries to predict |
| --- | --- | --- |
| Feed | Accounts you follow and suggested posts | How much time you will spend on the post or on the author's profile, whether you will send it in a direct message, or whether you will skip it |
| Stories | Accounts you follow | Whether you will tap into the story, reply, or move on to the next one |
| Reels | Accounts you follow and accounts you don't | Whether you will watch more of the reel than 95% of people who watched reels of the same length, reshare it, use its audio, follow the author, or leave before three seconds |
| Explore | Mostly accounts you don't follow | Whether you will like, save or share the post |

Instagram does not publish how much each prediction weighs or a formula you can reproduce. It does describe the groups of signals it uses: your activity in the app, information about the post (for example, how popular it is or when it was posted), information about who posted it, and your history with that account.

A common question is whether replying to comments on your posts counts among those signals. We go through what Instagram and TikTok say about comments in [does replying to comments boost reach on Instagram and TikTok?](/en/guides/does-replying-to-comments-boost-reach/).

One practical difference is worth keeping in mind. Feed and Stories mostly work with relationships that already exist. Reels and Explore are where a small account can reach people who don't follow it yet, and that is where predictions such as watching most of a video or following the author come into play.

## Followers and non-followers: two different diagnoses

Before you change the content, work out who you stopped reaching and measure it with views, because plenty of people watch a reel without liking or commenting. Since January 2025, each reel's insights split its views between followers and non-followers and include a view rate: the share of each group that keeps watching after the first three seconds, according to [TechCrunch](https://techcrunch.com/2025/01/29/instagram-gives-creators-more-insight-into-their-reels-performance/). For posts and stories, [account insights](https://help.instagram.com/788388387972460) split the accounts that engaged into followers and non-followers. Use interactions as a secondary signal.

If you're not sure what that metric counts or how it differs from the accounts that saw your content, we cover it in [what views on Instagram are](/en/guides/instagram-views/).

- **If it dropped among followers**, look at Feed and Stories. Those surfaces depend on your followers stopping, replying or sharing. Posting more of the same rarely fixes it.
- **If you barely reach non-followers**, look at Reels and Explore. First check that the content is eligible to be recommended; then check whether the start of the video holds attention.

## Before changing the video, check whether it can be recommended

Instagram applies [recommendation guidelines](https://about.instagram.com/blog/announcements/instagram-ranking-explained) in Explore, Reels, search and suggestions. A post can be allowed on Instagram and still not be eligible for recommendation. According to the 2023 explanation, **Account Status** is the best way to see whether your account and content can be recommended.

The same explanation notes that reels may get less visibility if they are low resolution or watermarked, muted, framed with borders, mostly text, or already posted on Instagram.

Originality matters too. In 2024, Instagram [announced](https://creators.instagram.com/recommendations-and-originality) that when it finds the original of reposted content, it shows the original instead of the copy in recommendations. [Meta's Latin American newsroom post about the same change](https://about.fb.com/ltam/news/2024/05/ayudando-a-los-creadores-a-encontrar-nuevas-audiencias/) (in Spanish, May 2, 2024) adds that accounts posting other people's content ten or more times in thirty days, without having created it or enhanced it in a material way, stop being shown where Instagram recommends content. They can become eligible again once thirty days have passed since their last unoriginal post, and Account Status shows that standing and lets them appeal. The rule doesn't affect publishers with licensing agreements or explicit permission from creators, nor what people who already follow those accounts see. Since November 17, 2026, eligible creators can also ask Instagram to alert them when someone copies their videos: see [Instagram content protection](/en/guides/instagram-content-protection/).

## Example: a reel that doesn't reach non-followers

Fictional case. A café posts a 20-second reel of latte art. It downloaded the video from its TikTok account, so it carries that watermark, and it opens with three seconds of logo. Almost all its views come from followers, and most of the few non-followers who see it leave before three seconds.

1. **Account Status.** No warnings: the account can be recommended, so the café starts by reviewing the video.
2. **The reel.** It has one issue Instagram documents, another app's watermark, and another that clashes with one of its predictions: an opening where nothing happens. The Reels system estimates whether someone will leave before three seconds, and a static logo gives them a reason to.
3. **The decision.** The café doesn't upload the same video again. It films a new take, exports it from the original file without a watermark, opens with the milk pouring into the cup, and moves the logo to the last second.
4. **How to assess it.** It compares non-follower views and their view rate with those of similar reels from the account, not with the best reel of the year. One reel doesn't prove the change worked; several posts made with the same approach can show a pattern.

Step 3 follows Adam Mosseri's advice not to repeat a reel that already worked and to [make a new version instead](/en/guides/adam-mosseri-repurpose-instagram-reels/). To see which part of a video loses viewers, our guide to [short-form video retention](/en/guides/short-form-video-retention/) shows how to read the curve.

## What you can change and what you can't

You can pick the format based on the surface you want to reach ([Reels or Stories](/en/guides/instagram-reels-vs-stories/)), post your own material, work on the opening of each video, and write captions that answer what someone would search for, as we explain in [Instagram and TikTok SEO](/en/guides/instagram-tiktok-search-optimization/).

You don't control each person's history or the weight of each signal, which Meta updates over time. Instagram also [states](https://about.instagram.com/blog/announcements/instagram-ranking-explained) that it doesn't suppress content to push accounts into buying ads. So the useful comparisons are between your own posts, over similar periods, with one change at a time, as explained in [content testing](/en/guides/social-media-content-testing/).

## Plan your next post in HeyMark.

Keep the idea, review the draft with your team, and see how it performed in the accounts you connected.

[Start free](https://app.heymark.ai) · [See how it works](https://heymark.ai/en/#product)

## Developer and agent resources

- [HeyMark MCP documentation](https://heymark.ai/en/mcp/)
- [llms.txt](https://heymark.ai/llms.txt)
- [Full site content for language models](https://heymark.ai/llms-full.txt)
- MCP protocol endpoint: `POST https://mcp.heymark.ai`
- OAuth protected-resource metadata: [/.well-known/oauth-protected-resource](https://mcp.heymark.ai/.well-known/oauth-protected-resource)
- MCP server card: [/server-card](https://mcp.heymark.ai/server-card)
