Speech is sensitive to mild cognitive changes due to age-related diseases, and prosodic features can identify patients with early-stage dementia from controls. Few studies have investigated speech markers of subtle cognitive impairment in non-neurodegenerative pathologies in younger populations, such as Cancer-Related Cognitive Impairment (CRCI). Little is known about the cognitive mechanisms underlying CRCI, but it is frequently encountered by cancer patients who mainly report memory-related concerns (i.e., forgetting words). Despite its substantial impact on patient quality of life, CRCI is difficult to detect with neuropsychological tools and often remains underdiagnosed. Our aim is to test whether previously documented speech markers are likely to detect CRCI in patients with breast cancer. We compared speech rate, F0 variability and pause duration in 11 breast cancer survivors with a cognitive complaint, 11 breast cancer survivors without any cognitive complaint and 10 controls in two narrative tasks (memory-based; picture-based). A Bayesian analysis showed no significant effects of group or task, but a qualitative analysis of pauses allowed us to generate hypotheses about the cognitive mechanisms underlying the patients’ reported memory concerns. Even though speech markers specific to CRCI have yet to be defined, prosodic analysis is a promising approach for detecting subtle cognitive impairment.