Psychological theories posit that we frequently switch social identities, yet little is known about the effectiveness of such switches. Our research aims to address this gap in knowledge by determining whether – and at what level of integration into the self-concept – a social identity switch impairs the activation of the currently active identity (“identity activation cost”). Based on the task-switching paradigm used to investigate task-set control, we prompted social identity switches and measured identity salience in a laboratory study using sequences of identity-related Implicit Association Tests (IATs). Pilot 1 (N = 24) and Study 1 (N = 64) used within-subjects designs with participants completing several social identity switches. The IAT congruency effect was no less robust after identity switches compared to identity repetitions, suggesting that social identity switches were highly effective. Study 2 (N = 48) addressed potential differences for switches between identities at different levels of integration into the self. We investigated whether switches between established identities are more effective than switches from a novel to an established identity. While response times showed the predicted trend towards a smaller IAT congruency effect after switching from a novel identity, we found a trend towards the opposite pattern for error rates. The registered study (N = 144) assessed these conflicting results with sufficient power and found no significant difference in the effectiveness of switching from novel as compared to established identities. An effect of cross-categorisation in the registered study was likely due to the requirement to learn individual stimuli.