TikTok Builds an Interest Map
A social network starts with people: whom you know, follow, or connect with. TikTok’s recommendation system starts more with content. It tries to infer which subjects, styles, and formats hold your attention, then uses that picture to choose what to show next.
The map is not a list of interests you fill out. It is inferred from behavior. Watching a video to the end, replaying it, moving on quickly, or interacting with it can all provide clues. Those signals help the system estimate whether a particular video—or similar videos—may suit you.
Each Feed Is a Feedback Loop
When TikTok recommends a video, your response becomes new evidence. If you linger on a cooking clip, the system may test other cooking videos; if you skip them, it has reason to adjust. This is a feedback loop: recommendations prompt behavior, and behavior reshapes recommendations.
The result can be discovery beyond a person’s existing circle. A creator’s video does not need to reach a viewer through a friend’s post to be considered for that viewer’s feed. The system is matching content to predicted interest, not simply passing along what a network of friends shares.
That approach makes the “For You” feed feel unusually personal, but it also means the feed reflects accumulated signals rather than a complete picture of what someone thinks or wants. A brief pause can be ambiguous; repeated viewing patterns give the system more to work with. The practical consequence is that small actions—especially what people keep watching—can influence what they encounter next.
