GeoMatch — Switzerland's AI-Assisted Refugee Resettlement Matching Pilot

GeoMatch is a machine-learning tool developed by researchers at the Stanford Immigration Policy Lab, in partnership with Switzerland's State Secretariat for Migration (SEM), to recommend which Swiss canton an asylum-seeking family should be resettled to in order to maximise their chances of finding employment. Since 2020, SEM has run the tool inside a randomized controlled trial: roughly half of arriving families receive an algorithmic recommendation for canton placement, while the rest are placed following standard practice.

Backtesting the algorithm on historical Swiss placement data — a 2013 cohort in which only 15% of resettled refugees found work within three years — researchers projected that algorithmic assignment could have raised that employment rate to around 26%, a 73–75% relative improvement, according to peer-reviewed methodology published by the Immigration Policy Lab. The pilot RCT has placed more than 1,000 refugee families to date.

This headline improvement figure comes from retrospective backtesting, not yet from the live government RCT itself. As of the most recent public reporting, three-year employment outcomes from the actual trial were still pending analysis, so the real-world impact of GeoMatch as deployed by SEM remains to be confirmed rather than established.