How do TikTok algorithms really work?

12 Ağu 07:23

Good questions, bro, because TikTok's algorithm is genuinely confusing. First off, the recommendations on the ForYou page aren't calculated "instantly"—they're constantly being updated. The system continuously gathers feedback based on user interactions (likes, comments, shares, watch time, scroll speed) and adjusts its suggestions accordingly. For example, if you watch a video for 3 seconds, don’t like it, and scroll away—at that moment, the algorithm registers it as "not interested" and starts showing you fewer similar videos. But the content you keep circling back to? That’s usually not because the algorithm thinks you’re "not interested"—it’s because you’re lingering on that content for short bursts. The system keeps learning, but sometimes those "lingering moments" can trick the algorithm. As for old videos getting recommended, well, there’s a kind of "priority window" built into the system. If a video doesn’t get high engagement quickly after being posted, it falls off the algorithm’s radar and can stay buried for months. But if an old video suddenly blows up because of a viral trend (like the #IceBucketChallenge), the algorithm kicks in right away and starts pushing it again. It’s kind of like those "windows" systems—those first few hours are critical. If you keep seeing the same content over and over, it’s usually not because the algorithm’s glitching—it’s more likely that you’ve got a thing for that content. Like, if you’re constantly watching videos under #CatsOfTikTok, the system labels you as a "cat lover" and just feeds you that stuff. In this case, it’s not the algorithm’s fault—it’s your own behavior doing the work. As for optimization, I think the best approach is to avoid sending signals that you’re not interested (i.e., scrolling away quickly from videos). Try to reduce the number of videos you watch for a few seconds and then swipe away, focus on watching content you actually like for longer periods, and engage with similar stuff. But remember, the algorithm isn’t 100% objective—it’s primarily designed to keep users on the platform as long as possible. So it’s less about "what you actually like" and more about "how can I keep you scrolling here?" That’s why you sometimes get weird recommendations, like seeing the same video again hours after you closed it. I mean, what can you do? It’s a system built around ad revenue, after all!