We consider a general model of a disordered copolymer with adsorption. This includes, as particular cases, a generalization of the copolymer at a selective interface introduced by Garel et al. [Europhys. Lett. 8 (1989) 9-13], pinning and wetting models in various dimensions, and the Poland-Scheraga model of DNA denaturation. We prove a new variational upper bound for the free energy via an estimation of noninteger moments of the partition function. As an application, we show that for strong disorder the quenched critical point differs from the annealed one, for example, if the disorder distribution is Gaussian. In particular, for pinning models with loop exponent 0 < α < 1/2 this implies the existence of a transition from weak to strong disorder. For the copolymer model, under a (restrictive) condition on the law of the underlying renewal, we show that the critical point coincides with the one predicted via renormalization group arguments in the theoretical physics literature. A stronger result holds for a "reduced wetting model" introduced by Bodineau and Giacomin [J. Statist. Phys. 117 (2004) 801-818]: without restrictions on the law of the underlying renewal, the critical point coincides with the corresponding renormalization group prediction. . This reprint differs from the original in pagination and typographic detail. 1 2 F. L. TONINELLI present a nontrivial localization-delocalization phase transition due to an energy-entropy competition.Mathematically, the model is defined in terms of a renewal sequence whose inter-arrival law has a power-like tail with exponent α + 1 ≥ 1. The model is exactly solvable in absence of disorder, and it turns out that the transition can be of any given order, from first to infinite, according to the value of α. This is therefore an ideal testing ground for physical arguments (Harris criterion, renormalization group computations) and predictions concerning the effect of disorder on the critical exponents and on the location of the critical point.The comprehension of this model has witnessed remarkable progress on the mathematical side, as proved by the recent book [17]. In particular it has been shown that for wetting, pinning or PS models, an arbitrary amount of disorder modifies the free-energy critical exponent if α > 1/2 [18], that is, disorder is relevant in this case, in agreement with the predictions of the so-called Harris criterion [23]. On the other hand, for 0 < α < 1/2 it has been proven recently [2,29] that if disorder is weak enough the free-energy critical exponent coincides with that of the homogeneous (i.e., nondisordered) model, and the (quenched) critical point coincides with the annealed one: disorder is irrelevant (again, in agreement with the Harris criterion). These results about "irrelevance" of disorder for 0 < α < 1/2 have been later refined and complemented in [21] with results about correlation-length critical exponents. The marginal case α = 1/2 is strongly debated in the theoretical physics literature: Ref. [14] claims that quenched and annealed c...