Shelterwood is commonly assumed to be a more nature-friendly silvicultural system than clear-cutting. However, dead-wood pools – a key characteristic of natural forest – have been seldom compared between these systems. We investigated how shelterwood harvesting influences the dynamics of different dead-wood fractions in Estonia, where the predominant forestry model is clear-cutting based but ‘seminatural’ (using native tree species and, to a significant extent, natural regeneration). We measured dead-wood pools in 49 Scots pine-dominated stands (representing all shelterwood harvesting stages), and in 11 pine-dominated and 10 Norway spruce-dominated stands as before-after experiments (1st stage only). We analysed dead-wood amounts in relation to site conditions and the proportion of timber harvested, and we compared the shelterwood impacts with published estimates from Estonian clear-cuts. Fine woody debris (5–9.9 cm) increased with the harvest. The volume of coarse woody debris was 19–27 m3 ha−1 in uniform shelterwood stands in pine forest (0–25 years after the first harvest); 63 m3 ha−1 in strip shelterwood stands in spruce forest (immediately post harvest). In before-after experiments, post-harvest dead-wood amounts depended on fraction and harvesting intensity, which determines the balance between the input of new debris (logs; stumps) and the loss of pre-existing standing and downed dead trees. After the first shelterwood harvesting, dead-wood pools remained relatively stable, which contrasts with the large fluctuations after clear-cutting. In the long term, however, shelterwood did not sustain generally larger dead-wood pools than the clear-cutting system in seminatural forestry setting. The issue to be resolved in both types of regeneration cuttings is the near-complete loss of standing dead trees, which probably requires new harvesting techniques.