Resonant atoms modify the electromagnetic environment that in turn determines their optical response, giving rise to a self-consistent light–matter problem at finite density. We investigate this feedback for the hydrogen Lyman-$\alpha$ transition using a local effective-medium description combined with an electromagnetic Green-function treatment of the atomic self-energy. Within the controlled regime, the self-consistent dressing produces a systematic density-dependent modification of the spectrum, including a shift of the spectral centroid and an asymmetric redistribution of integrated spectral weight. Comparison with a density expansion through fourth order provides an internal validity test: the local description remains quantitatively controlled up to approximately $2.0\times {10}^{21}{m}^{-3}$ and becomes marginal above this density. Formal continuation to higher densities produces much stronger spectral restructuring, but this occurs together with the loss of perturbative and local-response control and is therefore not interpreted as a quantitative prediction of resolved high-density modes. To assess experimental relevance without assuming a microscopic hydrogen collision model, we additionally introduce a parametric collisional broadening ${\Gamma }_{coll}=n\beta$. Collisional broadening becomes comparable to the controlled electromagnetic centroid shift for a characteristic coefficient of order ${10}^{-14}{m}^3{s}^{-1}$. The robust prediction of the local theory is thus a density-dependent electromagnetic dressing of the resonant response, whereas its high-density breakdown identifies the regime in which a spatially resolved, non-local microscopic description of collective electromagnetic coupling becomes necessary.