Immigration is a central issue in the political agendas of many Western parties, shaping the course of elections. Measuring population sentiments towards immigration is thus crucial to our understanding of the political outcomes the Western world is facing. Traditional methods for measuring these attitudes are often expensive—especially over time—and prone to biases, such as social desirability and recall errors. This project introduces a novel measure that leverages unobstructed behaviour to identify immigration preferences: user interactions with videos on TikTok. Using ‘reposts’ from a network of immigration-related videos, users are positioned on a latent spectrum of pro- vs anti-immigration ideology. A correspondence analysis computes users’ positions by minimising the distance with respect to videos using a weighted Euclidian metric. To test the proposed measure’s capacity to recover precise estimates of ideology for a given population, the aggregated score is compared against publicly available survey data on immigration attitudes in various Western countries. Finally, I further validate the measure by corroborating the position of political accounts with expert data and analysing the content of videos, to make sure the latent space is ideologically coherent. The main benefit of this methodology is that it can be applied to any dimension of interest for which a corpus of videos exists for a given time and country.