Personalized recommendations

Instagram builds personalized Feed recommendations by gathering public photos and videos that meet its quality and integrity rules, then ordering them according to their predicted value to the viewer.1 The information used includes previous engagement with similar content, interactions with its creator, and how many and how quickly other people like, comment on, share or save a post.13 Meta’s Newsroom explains that predicting whether someone will share a post is only one part of the process: its systems combine several predictions, including ones informed by surveys, because no single prediction fully measures value.5

Instagram’s personalized account suggestions match viewers with people to follow, using possibilities such as followers of accounts they already follow, accounts popular locally, and accounts related to topics or products they recently engaged with.2 The suggested-account system considers information such as mutual friends, removes blocked and deactivated accounts, and applies checks for Community Guidelines violations.2 For discovering posts rather than accounts, Meta’s engineering team described Instagram Explore in August 2023 as narrowing a large pool through several stages before adjusting the results for predicted engagement, content integrity and variety.6 Personal interests can also organize how recommendations are presented: Instagram’s announcement of its early TV test describes Reels grouped into channels matching a viewer’s interests, such as music, sports or travel.7 The audience need not be one viewer alone: Mosseri describes Blend as a shared Reels feed that combines recommendations friends may like.8

Instagram’s Help Center says its personalization controls can show a summary of a viewer’s recent interests; the feature is still in testing and may not be available to everyone, and the parts of Instagram affected can vary by region.9 For a broader fresh start, Mosseri described a recommendation-reset test for Explore, Reels and Feed in November 2024; after a reset, suggestions would gradually become personalized again through interactions with content and accounts.10 Instagram for Creators also describes controls for marking recommendations as uninteresting and adjusting how much sensitive content someone is comfortable seeing.3 These preferences operate within recommendation standards: Meta’s Newsroom explains that some content may remain allowed on its platforms without qualifying to appear in recommendations.11