Whether participation in active labour market programmes (ALMPs) pushes individuals back into employment depends on the programme's characteristics. On the basis of encompassing registry data that allow us to control for usually unobserved employability, we find evidence of a systematic access bias whereby jobcentre caseworkers in Switzerland assign unemployed persons to activation measures based on a competition logic that is mainly driven by an economic rationale and the jobcentre's performance evaluation. This practice seems problematic because it results in an overrepresentation of immigrants in measures with little efficacy rather than in measures that could compensate for employability disadvantages. Conversely, Swiss citizens are more likely to enter beneficial human‐capital‐intensive measures. It is plausible that this discrepancy in programme participation amplifies the general labour market disadvantages of immigrants.