Rasero
A fixed, public method to measure the politics social feeds serve to new accounts.
Visit rasero.org ↗The website itself is in Spanish.

Rasero measures what political content social networks serve to a brand-new account in Spain: how much arrives, with what leaning and how soon it appears. It focuses on the networks where people usually open their first account (TikTok, YouTube Shorts, Instagram and X), and the method and classification manual were published before any data was collected, with a dated change log.
Every session follows the same protocol on every network: 25 minutes, 8 seconds per item whatever its content, and no interaction. Capture runs in an automated browser on a home connection, audio is transcribed on the same computer, and a language model classifies each item blind, without knowing which network it came from, using rules that have to work the same way for both sides. The project carries no advertising and has no funding.
What makes it special
- 01Method and classification manual published before measuring, with a dated change log.
- 02A trial of the manual with 677 items from TikTok and YouTube Shorts over 4 sessions, published with its limitations.
- 03Four classification axes (political or not, leaning, traits and tone) and a symmetry test for every rule.
- 04A review of 15 earlier studies and projects on political bias in recommendation algorithms.
