Instagram is making efforts to clarify how its recommendation system works and dispel misconceptions surrounding its algorithm and the practice of shadowbanning. In a blog post by Adam Mosseri, Instagram’s top executive, he explains that the app does not have a single algorithm governing what users see. Instead, different algorithms and ranking systems are utilized for different sections of the app, such as Explore, Reels, Stories, and search. These algorithms consider various signals to determine how content is ranked for each user.
Mosseri provides examples of how content is ranked in different sections. The main feed order is based on users’ past activity and interactions with the creators. Stories take into account viewing history and the closeness between users. Explore recommendations are primarily influenced by posts users have liked, saved, shared, and commented on, with a possibility of including content from accounts users haven’t interacted with before.
The blog post also addresses the concept of shadowbanning, acknowledging that the term lacks a universal definition but is often used to imply that an account or content is limited or hidden without a clear explanation. Instagram is working on increasing transparency around cases where creators’ content or accounts are blocked from recommendations. Mosseri highlights the “account status” feature, which alerts users if their posts or account are deemed ineligible for recommendations and provides an appeals process.
Instagram is striving to enhance transparency when a creator’s content is not widely distributed due to policy violations. The company is testing new notifications to inform creators when the reach of their Reels may be limited due to watermarks. The details shared in the blog post highlight the increasing importance of algorithmic recommendations within Instagram. While there may be some creators who find the explanations unsatisfying, Instagram’s aim is to transform into a “discovery engine” focused on recommendations rather than just friends’ posts.