Context. Direct imaging and spectroscopy of Earth-like planets and young Jupiters require contrast values up to 10 6 − 10 10 at angular separations of a few dozen milliarcseconds. To achieve this goal, one of the most promising approaches consists of using large segmented primary mirror telescopes with coronagraphic instruments. While they are able to reach high contrast at small angular separations, coronagraphs are highly sensitive to wavefront errors, however. The segmentation itself is responsible for phasing errors and segment-level vibrations that have to be controlled at a subnanometric accuracy. Aims. We propose an innovative method for a coronagraph design that allows a consequent relaxation of the segment phasing and stability constraints for low segment-count mirrors and generates an instrument that is more robust to segment-level wavefront errors. Methods. This method is based on an optimization of the coronagraph design that includes a segment-level apodization. This is repeated over the pupil to match the segmentation redundancy and improves the contrast stability beyond the minimum separation set by the single-segment diffraction limit. Results. We validate this method on a Giant Magellan Telescope (GMT)-like pupil (seven circular segments) for two coronagraph types: apodized pupil Lyot coronagraphs, and apodizing phase plate coronagraphs. For the apodized pupil Lyot coronagraphs, redundant apodization enables releasing the piston phasing constraints by a factor of 5 to 20 compared to classical designs. For the apodizing phase plate coronagraphs, the contrast remains almost constant up to 1 radian RMS of the phasing errors. We also show that redundant apodizations increase the robustness of the coronagraph to segment tip-tilt errors, as well as to missing segment errors. Conclusions. Redundant apodization enables reducing or even removing any constraints on the primary mirror segment phasing at the price of larger angular separations and lower throughputs. This method cannot be applied to higher-segment count mirrors such as the ELT or the TMT, but it is particularly suitable for low segment-count mirrors (fewer than ∼ 20 segments) such as the GMT aperture. These mirrors aim for high-contrast imaging of debris disks or exoplanets down to 100 mas.