The Tackling Hate Lab has published a new report titled The 2024 UK Riots: Tracing the Path from Online Activity to Offline Mobilisation. The report asks a simple but important question: can online hate and political discourse signal, or even anticipate, real-world violence?
The study focuses on the riots that followed the Southport attack in July 2024. It uses three datasets: 67,934 geolocated tweets from the UK, 5,388 from Australia as a comparison case, and 85 verified offline incidents across 37 locations. Using AI-based classifiers, the team identifies two key forms of online discourse: anti-foreigner and anti-politics language, and tracks how these evolve over time and space.
First, spikes in online hate align closely with periods of offline unrest, but raw tweet volume alone does not predict violence. Second, original posts containing hostile narratives increase before riots, while retweets peak during the violence, suggesting a shift from mobilisation to amplification. Third, reposting rates, that is how widely each message spreads, emerge as a strong early warning signal. Finally, a small number of highly connected users drive most engagement, and coordinated clusters sustain pro-violence narratives.
The report shows that not all online hate matters equally. Content that mobilises and spreads widely is what signals risk.