User-content affinity

Instagram’s process for matching suggested posts to viewers begins with a selection of public photos and videos that meet its quality and integrity rules.2 Meta’s Transparency Center says the system then considers information such as how the viewer has engaged with similar content.2 Its listed inputs also include posts the viewer previously reshared and how often recommended content has been displayed to them.2

In Meta’s patents, affinity scores are numerical estimates of relationship strength or interest; all patent material discussed here describes proposals, not confirmed Instagram implementation.3456 Meta’s patents describe affinity broadly enough to include a person’s interest in content, making a viewer–post match one application of the concept.346 Meta’s September 2022 patent proposes storing these scores alongside information about connections between a person and other people, interests or items.5 The same Meta patent says the system may compute affinity scores over time, rather than describing them solely as a one-time calculation.5

Meta’s December 2023 patent gives a concrete proposed use of viewer–content matching for posts featuring products.7 The Meta patent proposes that, after a viewer selects a product, the system can show other posts featuring it according to how well those posts match that viewer.7 The same Meta patent proposes providing product insights based on predicted user engagement, applying the matching concept to estimates of which products might attract a response.7

Instagram for Creators places viewer–content matching alongside observed engagement when explaining how widely a post is shown.1 Its FAQ identifies engagement per viewer as an important consideration, including watch time, sharing, following, liking, commenting and profile visits.1 It says techniques for matching content to interested audiences are used in some scenarios, rather than describing them as part of every recommendation.1 Its practical example is to consider themes that previously reached the intended audience, whether posted by the creator or another creator in the same niche.1