Abstract In higher education systems with no or low entry requirements for domestic students, post-entry language assessments (PELAs) are often used to detect at-risk students. This study investigates how two language proficiency measures relate to students’ academic achievement: an academic reading and vocabulary (ARV) screening test, and the computer test of the Interuniversitaire Taaltest Nederlands voor Anderstaligen (ITNA), a Dutch L2-entrance test. The ARV screening test, ITNA and its language in use subtest show moderate to weak correlations with achievement. While the ARV screening test selects a large potential at-risk group; the few students failing the ITNA, language in use or scoring lower than ‘very good’ on language in use have a higher risk of underperforming. When controlling for students’ background variables, only academic language proficiency remains a significant predictor. While academic language proficiency seems more appropriate as a PELA construct, general language proficiency may identify a small, but high-risk group.