TikTok adds a Refresh button to For You, resetting the recommendation algorithm to show videos like for a new user, after testing the feature in February 2023
Mia Sato / The Verge :
Context & Ripple Effects
The Refresh button completes a transparency arc TikTok started last winter, when it began telling users why a video was in their For You feed — citing watches, likes, shares, searches, and region. Explanations said what shaped a feed; the reset lets users discard it entirely and return to a new-signup state after February testing.
The timing matters because TikTok's recommender has been under external scrutiny — Global Witness found substantial far-right bias in For You recommendations ahead of the German elections — so handing users an escape hatch is also a governance gesture. The pattern proved exportable: over a year later, Instagram was still following the template, testing resets of the algorithmic suggestions powering Feed, Reels, and Explore.
First-order effects
- TikTok users dissatisfied with their For You feed no longer need to manually unfollow, mute, or grind through weeks of 'not interested' taps — one tap wipes the learned profile, at the cost of losing a feed tuned to their tastes.
- Creators who built audiences on entrenched recommendation patterns now face feeds where a reset removes them from users' learned profiles, making retention dependent on re-earning distribution rather than accumulated signal.
Second-order effects
- Rivals get a copyable feature spec: Instagram's eventual test of Feed, Reels, and Explore resets shows the reset button moving from differentiator to table stakes among algorithmic-feed platforms.
Third-order effects
- If resets plus recommendation explanations become standard controls, algorithmic curation shifts from an opaque black box platforms defend toward a governed surface users can inspect and clear — with regulators likely to treat such user-facing controls as the baseline expectation rather than a goodwill gesture.
The trend: Recommendation algorithms are being opened to user control — explanations first, then full resets — as platforms convert algorithmic opacity from a moat into a managed liability.