Within the Ginzburg-Landau theory, which is quantitatively correct for classical superconductors, it is shown that a type-I superconductor can display an interface delocalization or ''wetting'' transition, in which a macroscopically thick superconducting layer intrudes from the surface into the bulk normal phase. The condition for this transition to occur is that the superconducting order parameter \psi\(2) is enhanced at the surface. This corresponds to a negative surface extrapolation length b. The wetting transition takes place at built two-phase coexistence of normal and superconducting phases, at a temperature T-D below the critical temperature T-c, and at magnetic field H-D = H-c(TD). The field is applied parallel to the surface. Surprisingly, the order of the wetting transition is controlled by a bulk material constant, the Ginzburg-Landau parameter kappa. This is very unusual, since in other systems (fluids, Ising magnets,...) the order of the wetting transition depends on surface parameters that are difficult to determine or control. For superconductors, first-order wetting is predicted for 0 less than or equal to kappa<0.374, and critical wetting for 0.374<kappa<1/root 2. In the case of first-order wetting, the prewetting extension is also found. Unlike in standard wetting problems, the prewetting line does not terminate at a critical point but changes from first to second order at a tricritical point. Twinning-plane superconductivity (TPS) is reinterpreted as a prewetting phenomenon. The possibility of critical wetting in superconductors is especially interesting because this phenomenon has largely eluded experimental verification in any system until now. Furthermore, superconductors provide a realization of wetting in systems with short-range (exponentially decaying) interactions. This is very different from the usual long-range (algebraically decaying) interactions, such as van der Waals forces, and has important consequences for the wetting characteristics.