We investigated individual differences in faking in simulated high-stakes personality assessments through the lens of expectancy (VIE) theory, using a novel experimental paradigm. Three hundred ninety-eight participants (MTurk) completed a “low-stakes” HEXACO personality assessment for research purposes. Three months later, we invited all 398 participants to compete for an opportunity to complete a genuine, well-paid, one-off MTurk job, and 201 accepted. After viewing the selection criteria, which described high levels of perfectionism as critical for selection, these participants completed the HEXACO personality assessment as part of their applications (“high-stakes”). All 201 participants were then informed their applications were successful and were invited to complete the performance task, with 189 accepting the offer. The task, which involved checking text data for inconsistencies, captured two objective performance criteria. We observed faking on measures of diligence and perfectionism. We found that perceived job desirability (valence) was the strongest (positive) determinant of individual differences in faking, along with perceived instrumentality and expectancy. Honesty-humility was also associated with faking however, unexpectedly, the association was positive. When all predictors were combined, only perceived job desirability remained a significant motivational determinant of faking, with cognitive ability also being a positive predictor. We found no evidence that cognitive ability moderated the relations of motivation and faking. To investigate the role of faking on predictive validity, we split the sample into those who had faked to a statistically large extent, and those who had not. We found that the validity of high-stakes assessments was higher amongst the group that had faked.