Socially desirable responding can impair the validity of self-report questionnaires, especially in high-stakes situations in which people are incentivized to manage the impression they make on others. The current experiment examined the context dependency of impression management. Participants (N = 231) completed the Big Five Inventory-2 twice, first honestly and then with faking-good instructions in a job or dating context. Socially desirable responding was present in both contexts but was more pronounced in the job context than in the dating context for many (but not all) Big Five domains and facets. Future research should investigate whether faking behavior differs across contexts not only under faking-good instructions but also in high-stakes situations (e.g., personnel selection or online dating).