Objectives: Older adults with generalized anxiety disorder (GAD) have elevated diurnal cortisol patterns and show an increased cortisol stress response, which may increase risk for cognitive dysfunction. The current secondary data analysis examined how neuropsychological assessment as a possible laboratory stressor affects cortisol levels in late-life GAD and, in turn, how cortisol levels affect cognitive performance. Methods: The current sample consisted of 69 individuals with late-life GAD and 39 psychiatrically healthy group-matched comparison participants. Cognitive performance was measured with a neuropsychological battery and salivary cortisol was collected at several time points. Hierarchical regressions were performed to assess the moderating role of cortisol in the relationship between GAD status and cognitive performance. Results: The results revealed that older adults with GAD showed significantly lower cortisol levels during neuropsychological assessment, compared to their baseline levels. Further, there was a significant interaction between post-neuropsychological assessment cortisol levels and GAD status on several measures of cognitive performance. The interaction indicated that there is a significant negative relationship between cortisol level and cognitive performance in the GAD participants and no such relationship in the comparison participants. Conclusions: Our results revealed that participating in a neuropsychological assessment was associated with reduced cortisol in GAD participants, suggesting that refocusing attention such as engaging in cognitive tasks had a cortisol-lowering effect. Further, a higher cortisol level appears to have a detrimental effect on cognitive performance for individuals with GAD, but not psychiatrically healthy comparison participants. The methodological and treatment implications of these findings are discussed.