It has been widely assessed that very preterm children (<32 weeks gestational age) present language and memory impairments compared with full-term children. However, differences in their underlying semantic memory structure have not been studied yet. Nevertheless, the way concepts are learned and organized across development relates to children's capacities in retrieving and using information later. Therefore, the semantic memory organization could underlie several cognitive deficits existing in very preterm children. Computational mathematical models offer the possibility to characterize semantic networks through three coefficients calculated on spoken language: average shortest path length (i.e., distance between concepts), clustering (i.e., local interconnectivity), and modularity (i.e., compartmentalization into small sub-networks). Here we assessed these coefficients in 38 very preterm schoolchildren (aged 8-10 years) compared with 38 full-term schoolchildren (aged 7-10 years) based on a verbal fluency task. Using semantic network analysis, very preterm children showed a longer distance between concepts and a lower interconnectivity at a local level than full-term children. In addition, we found a trend for a higher modularity at a global in very preterm children compared with full-term children. These findings provide preliminary evidence that very preterm children demonstrate subtle impairments in the organization of their semantic network, encouraging the adaptation of the support and education they receive.