Past studies have reported divergent results regarding the effect of mobile devices on general mental ability (GMA) test scores. We investigate selection bias as an explanation for this inconsistency in GMA score differences between applicants using mobile or nonmobile devices reported in observational and lab studies. We initially found that mobile test-takers scored 0.58 SD lower than nonmobile test-takers in an operational sample of 76,948 applicants across over 400 occupations. However, we found that mobile device use was more prevalent among applicants with lower educational attainment and within jobs of lower complexity. These factors, among others, could potentially confound the observed GMA score differences between devices. The device effect shrank to d = 0.25 after controlling for selection bias in device choice using propensity score weighing. As an alternative, we also used poststratification to control for selection bias and this yielded an even weaker device effect (d = 0.10). Our results indicate that the large device effects obtained in prior operational studies are possibly inflated by selection bias. Therefore, it is important to control for these demographic and occupational differences between self-selected device groups when analyzing operational data for research purposes. Propensity score weighing and poststratification appear useful for reducing the impact of selection bias in real-world, observational data. We also strongly recommend the use of random assignment to prevent selection bias when evaluating device effects for new or adapted GMA or similar ability tests. (PsycInfo Database Record (c) 2023 APA, all rights reserved).