In the neonatal intensive care unit, there is a lack of understanding about how best to communicate the prognosis of a serious complication to parents. To examine parental preferences and the effects of optimistic vs pessimistic message framing when providing prognostic information about a serious complication. This crossover randomized clinical trial was conducted at a single German university medical center between June and October 2021. Eligible participants were parents of surviving preterm infants with a birth weight under 1500 g. Data were analyzed between October 2021 and August 2022. Alternating exposure to 2 scripted video vignettes showing a standardized conversation between a neonatologist and parents, portrayed by professional actors, about the prognosis of a hypothetical very preterm infant with severe intraventricular hemorrhage. The video vignettes differed in the framing of identical numerical outcome estimates as either probability of survival and probability of nonimpairment (optimistic framing) or a risk of death and impaired survival (pessimistic framing). The primary outcome was preference odds (ratio of preference for optimistic vs pessimistic framing). Secondary outcomes included state anxiety, perceptions of communication, and recall of numerical estimates. Of 220 enrolled parents (142 [64.5%] mothers; mean [SD] age: mothers, 39.1 [5.6] years; fathers, 42.7 [6.9] years), 196 (89.1%) preferred optimistic and 24 (10.1%) preferred pessimistic framing (preference odds, 11.0; 95% CI, 6.28-19.10; P < .001). Preference for optimistic framing was more pronounced when presented second than when presented first (preference odds, 5.41; 95% CI, 1.77-16.48; P = .003). State anxiety scores were similar in both groups at baseline (mean difference, -0.34; -1.18 to 0.49; P = .42) and increased equally after the first video (mean difference, -0.55; 95% CI, -1.79 to 0.69; P = .39). After the second video, state anxiety scores decreased when optimistic framing followed pessimistic framing but remained unchanged when pessimistic framing followed optimistic framing (mean difference, 2.15; 95% CI, 0.91 to 3.39; P < .001). With optimistic framing, participants recalled numerical estimates more accurately for survival (odds ratio, 4.00; 95% CI, 1.64-9.79; P = .002) but not for impairment (odds ratio, 1.50; 95% CI, 0.85-2.63; P = .16). When given prognostic information about a serious complication, parents of very preterm infants may prefer optimistic framing. Optimistic framing may lead to more realistic expectations for survival, but not for impairment. German Clinical Trials Register (DRKS): DRKS00024466.