We study the effect of radiogenic heating due to 26Al on the thermal evolution of small icy satellites. Our object is to find the extent of internal melting as a function of the satellite radius and of the initial 26Al abundance. The implicit assumption, based on observations of young stars, is that planet and satellite accretion occurred on a time scale of approximately 10(6) yr (comparable with the lifetime of 26Al). The icy satellites are modeled as spheres of initially amorphous ice, with chondritic abundances of 40K, 232Th, 235U, 238U, corresponding to an ice/dust mass ratio of 1. Evolutionary calculations are carried out, spanning 4.5 x 10(9) yr, for different combinations of the two free parameters. Heat transfer by subsolidus convection is neglected for these small satellites. Our main conclusion is that the initial 26Al abundance capable of melting icy bodies of satellite size to a significant extent is more than 10 times lower than that prevailing in the interstellar medium (or that inferred from the Ca-Al rich inclusions of the Allende meteorite, approximately 7 x 10(-7) by mass). We find, for example, that an initial 26Al mass fraction of approximately 4 x 10(-8) is sufficient for melting almost completely icy spheres with radii of 800 km, typical of the larger icy planetary satellites. We also find that for any given 26Al abundance, there is a narrow range of radii below which only marginal melting occurs and above which most of the ice melts (and refreezes later). Since extensive melting may have important consequences, such as differentiation, gas release, and volcanic activity, the effect of 26Al should be included in future studies of satellite interiors.