User interests

Instagram’s Help Center says suggested posts depend partly on accounts someone follows and posts they have liked, saved, reposted or commented on; Instagram’s ranking explainer also identifies recent engagement as a clue26. Its Help Center gives a concrete example: interacting with restaurants and bookstores can lead to recommendations about food, recipes, books or reading3. Instagram for Creators says recommendations also consider a person’s interaction history with the account posting and how other people respond to the post4. A Meta patent about location-based recommendations proposes organizing inferred interests into broad or specific categories—for example, treating a liked shoe article as interest in a brand, shoes or clothing—but does not establish that Instagram uses that mechanism8.

Instagram for Creators says Feed, Stories, Explore and Reels each use their own system for choosing and ordering content, tailored to how people use that part of the app9. Instagram’s Reels explainer describes predictions about watching to the end, liking, finding a video entertaining or visiting its audio page, alongside surveys about whether a reel was entertaining or funny10. Meta’s Newsroom says no single prediction perfectly measures a post’s value to someone, so its Facebook and Instagram systems combine multiple predictions based on behavior and survey feedback7.

Mosseri illustrates shared interests with someone sending a soccer highlight to a friend who also loves soccer12. He has also described showing people Reels their friends liked so they can recognize a shared enthusiasm, such as comedy or architecture13. People can then start a conversation with the friend whose like they saw, making the content a starting point for interaction rather than only something to watch13.

Instagram’s Help Center includes Accounts You May Like among the recommendations intended to help people discover relevant communities and content3. A Meta patent proposes identifying accounts as authorities on a topic through the interests of their followers, then using that assessment to recommend accounts14. Another Meta patent proposes recommending groups organized around common interests and highlighting members a person knows or has a connection with; these patent proposals are not confirmation of Instagram’s account- or group-recommendation methods1415.

Mosseri’s demonstration of Your Algorithm included labels such as fashion, award-show glamour, “sweat sessions” and “culinary indulgences”21. He clarified that “sweat sessions” referred to workouts, illustrating the wording people may encounter when reading their inferred interests21. Meta’s Newsroom says parents and guardians using supervision can view the general interest categories their teens engage with, giving them broader visibility into the kinds of content their teens see18.

Meta’s Newsroom says Hidden Words can hide recommended content containing specified words or phrases in captions1. A Meta patent describes explicit profile declarations such as favorite books, music or an interest in photography as information for assessing relevance22. Another Meta patent describes selecting a football or team icon on an event-related post to express an interest; this is a patent example of declaring a preference, not an Instagram instruction23.

Instagram’s Help Center says ads can reflect followed accounts, liked posts, content someone makes or interacts with, activity on third-party websites and apps, and feedback showing interest or lack of interest in an ad24. It also lists Facebook information and interests, and says activity on Facebook may affect Instagram ads when the accounts are linked24. Depending on the region, interactions with AI products may also contribute24. For suggested products, Instagram’s Help Center says activity across Meta products can influence what appears in Feed, Reels and Explore25.

Mosseri said in October 2025 that ads or product suggestions can partly reflect friends’ interests and the interests of other people with similar tastes26. He explained that an apparently surprising ad might follow a friend’s earlier search for a product, or reflect searches by people with similar interests26. He rejected microphone listening as the explanation and also identified unnoticed earlier exposure to an ad and coincidence as possibilities26.

A Meta patent proposes detecting product-category interest when related activity reaches a specified level, then presenting personalized recommendations, ratings and reviews27. Another Meta patent proposes ordering paid stories alongside ordinary posts using both the viewer’s interest and the advertiser’s payment28. A Meta patent concerning life-event messages proposes gift suggestions based partly on the recipient’s interests and merchant bids; these commercial patent proposals do not establish which methods Instagram uses272829.

Mosseri said in August 2025 that creators do not need to interact with content similar to their own to maximize reach30. He encouraged creators to interact with the content they personally love30. His advice for publishing was to make material they think their audience will find compelling30.

Mosseri says people actively unfollow accounts to make Instagram more interesting to them, particularly younger users31. He explains that audiences can grow out of subjects: someone attracted to content aimed at a younger audience may later move on to something else31. He also says reduced interest is only one possible reason followers do not see a post, alongside limited scrolling, mistakes in how Instagram orders posts and safety measures32. A follower becoming less interested therefore does not establish that the creator’s content is bad32.