Instagram head Adam Mosseri says originality, recency, discovery beyond an account’s existing followers and larger AI models all shape ranking alongside predicted engagement. His latest ranking series explains how Instagram thinks about distribution, rather than announcing a new algorithm update or a fixed list of signal weights.
In a series of videos on Instagram, Mosseri has been explaining why the app ranks content and what its systems try to predict. Those predictions include whether someone is likely to watch, like or share a post, while broader priorities such as originality and recency influence what gets surfaced.
For creators, the useful distinction is that Instagram is not describing one universal formula. Different recommendation systems make predictions from multiple signals, and Mosseri’s latest explanation does not assign public weights to them.
Why Instagram ranks content at all
Mosseri argued that a purely unranked feed would favour accounts that post most often. Brands and organisations can publish far more frequently than an average user, he said, which can push posts from friends and creators further down the feed. Instagram instead ranks content by predicting how likely someone is to watch, like, share or otherwise respond to it.
Mosseri points creators towards engagement rate
Mosseri said creators should look at engagement relative to reach rather than relying only on raw views or likes. A post seen by a smaller audience can still produce a stronger response rate than one with a larger view count, making the ratio more useful when comparing performance across posts.
Four priorities beyond engagement
Mosseri highlighted four broader priorities in Instagram’s ranking direction. They should not be read as four equally weighted ranking signals.
- Originality:
Instagram says it tries to give more distribution to original content and the creator who produced it, rather than accounts reposting the same material.
- Recency:
When something is posted matters. Mosseri said recency remains important enough that the feed is still roughly chronological, even though it is ranked.
- Breaking through:
Instagram can test content from public accounts with people who do not already follow them. If the response is strong, distribution can widen, giving smaller or unfamiliar creators a route into non-follower discovery.
- Larger AI models:
Instagram is using larger AI models in its recommendation systems so they can process more information when predicting what a person may want to see. Mosseri was referring to ranking infrastructure, not to a preference for AI-generated content.
Trial Reels sit outside the core ranking principles
Mosseri separately encouraged creators to keep experimenting with formats and topics, pointing to Trial Reels as one way to test ideas. Trial Reels are shown to non-followers first, allowing creators to gauge response before deciding whether to share them more broadly. This was presented as a testing tool, not as another ranking signal.
Like this story? get the next one in your inbox.
Users can also push recommendations in a different direction
Mosseri also pointed to Favourites, Not interested and Your Algorithm as ways users can influence what Instagram recommends. Your Algorithm lets people add or remove topics that Instagram has associated with their interests.
The series does not reveal a formula creators can optimise against. It instead reinforces a few established priorities: Instagram predicts audience response, favours original and recent content, gives eligible public posts a chance to reach non-followers and is investing in larger recommendation models. The exact weighting of those factors remains undisclosed.
