The semantic fluency test is one of the most widely used neuropsychological tests in dementia diagnosis. Research utilizing the qualitative, psycholinguistic information embedded in its output is currently underexplored in presymptomatic and prodromal genetic FTD. Presymptomatic MAPT (n = 20) and GRN (n = 43) mutation carriers, and controls (n = 55) underwent up to 6years of neuropsychological assessment, including the semantic fluency test. Ten mutation carriers became symptomatic (phenoconverters). Total score and five qualitative fluency measures (lexical frequency, age of acquisition, number of clusters, cluster size, number of switches) were calculated. We used multilevel linear regression modeling to investigate longitudinal decline. We assessed the co-correlation of the qualitative measures at each time point with principal component analysis. We explored associations with cognitive decline and grey matter atrophy using partial correlations, and investigated classification abilities using binary logistic regression. The interrater reliability of the qualitative measures was good (ICC = 0.75-0.90). There was strong co-correlation between lexical frequency and age of acquisition, and between clustering and switching. At least 4years pre-phenoconversion, GRN phenoconverters had fewer but larger clusters (p < 0.001), and fewer switches (p = 0.004), correlating with lower executive function (r = 0.87-0.98). Fewer switches was predictive of phenoconversion, correctly classifying 90.3%. Starting at least 4years pre-phenoconversion, MAPT phenoconverters demonstrated an increase in lexical frequency (p = 0.009) and a decline in age of acquisition (p = 0.034), correlating with lower semantic processing (r = 0.90). Smaller cluster size was predictive of phenoconversion, correctly classifying 89.3%. Increase in lexical frequency and decline in age of acquisition were associated with grey matter volume loss of predominantly temporal areas, while decline in the number of clusters, cluster size, and switches correlated with grey matter volume loss of predominantly frontal areas. Qualitative aspects of semantic fluency could give insight into the underlying mechanisms as to why the "traditional" total score declines in the different FTD mutations. However, the qualitative measures currently demonstrate more fluctuation than the total score, the measure that seems to most reliably deteriorate with time. Replication in a larger sample of FTD phenoconverters is warranted to identify if qualitative measures could be sensitive cognitive biomarkers to identify and track mutation carriers converting to the symptomatic stage of FTD.