Abstract Power-law size distributions are the hallmarks of nonlinear energy dissipation processes governed by self-organized criticality (SOC). Here we analyze 75 data sets of stellar flare size distributions, mostly obtained from the Extreme-Ultraviolet Explorer and the Kepler mission. We aim to answer the following questions for size distributions of stellar flares. (i) What are the values and uncertainties of power-law slopes? (ii) Do power-law slopes vary with time? (iii) Do power-law slopes depend on the stellar spectral type? (iv) Are they compatible with solar flares? (v) Are they consistent with SOC models? We find that the observed size distributions of stellar flare fluences (or energies) exhibit power-law slopes of α E = 2.09 ± 0.24 for optical data sets observed with Kepler. The observed power-law slopes do not show much time variability and do not depend on the stellar spectral type (M, K, G, F, A, giants). In solar flares, we find that background subtraction lowers the uncorrected value of α E = 2.20 ± 0.22 to α E = 1.57 ± 0.19. Furthermore, most of the stellar flares are temporally not resolved in low-cadence (30 minutes) Kepler data, which causes an additional bias. Taking these two biases into account, the stellar flare data sets are consistent with the theoretical prediction of SOC models, i.e., α E = 1.5. Thus, accurate power-law fits require automated detection of the inertial range and background subtraction, which can be modeled with the generalized Pareto distribution, finite-system size effects, and extreme event outliers.