The relationship between self-estimated and observed intelligence has sparked an interest among researchers in recent years. However, studies on the relationship between how personality traits predict self-estimations of intelligence in different test-settings are still not explored. This study set up a two-way experimental within-subjects design (N = 84) to test how prior personality traits predict self-estimated intelligence, immidiately after having performed an intelligence-test. At a first occasion, an explicitly easy intelligence-test (Kajonius, 2014) under relaxed pretenses was given to a number of work psychology students, and at a second occasion two months later a difficult intelligence test (Condon & Revelle, 2014, ICAR16) with time pressure was given to the same students. No test-scores or personality traits’ results were revealed until after the study. An expected relationship between self-estimated and observed intelligence of similar strength was found in both conditions (r ~ .55). Also, the students’ prior personality tests (Big Five Inventory-44) showed that neuroticism had a negative relationship with self-estimated intelligence (r ~ -.30), similarly in both conditions. However, the self-enhancing traits of Machiavellianism and narcissism (Short Dark Triad-3) showed a significant positive relationship with self-estimated intelligence only in the easy condition (r = .30) and not in the difficult condition (r = .10). The discussion suggests that neither student exits nor adjustment effects can explain the differences in the impact of dark triad traits between the conditions. The conclusion is that the study implies that test situations can activate personality traits differently, while the accuracy of self-perceived intelligence is more constant.