TikTok's US Algorithm Gets Retrained

January 28. TikTok's US entity is retraining its recommendation engine under new ownership. The app's algorithm, the one that decides what shows up on the For You page, is being rebuilt from scratch.

The mechanics matter here. An algorithm isn't neutral. It's a set of decisions baked into code: what gets weighted, what gets suppressed, what signals count. The old TikTok algorithm ran on Chinese servers and optimized for engagement metrics built in Shenzhen. The new one runs on US infrastructure and optimizes for metrics built in Mountain View or Seattle or wherever the new parent company's engineering team sits.

What changes when the weights change? Surface-level: probably nothing dramatic in week one. A creator who was doing well on the old algorithm might notice their reach plateaus. A brand that had cracked the code on TikTok Shop might see conversion rates shift. The feed still scrolls, the videos still play, but the invisible hand sorting them has different priorities now.

The retraining process itself is the tell. You don't retrain an algorithm unless the old one was serving goals that don't align with the new owner's goals. The old TikTok wanted maximum time-on-app and maximum virality, because virality outside the US didn't threaten the parent company. The new TikTok wants those things too, but also wants to avoid the regulatory scrutiny that comes with being accused of pushing harmful content, foreign influence, or data privacy violations.

So the new algorithm will likely be more conservative. More brand-safe. More aligned with what US advertisers want to see next to their ads. Which means: less chaotic, less weird, less of the raw unfiltered feed energy that made TikTok TikTok in the first place.

The trial mentioned in the same breath as this news, the one claiming social platforms are addictive by design, isn't adjacent to this story. It's the same story. The algorithm is the addiction mechanism. Retraining it is retuning the dopamine loop. Whether the new loop is less addictive or just addictive in a different way is the question no one's answering yet.