When the copula of the conditional distribution of two random variables given a covariate does not depend on the value of the covariate, two conflicting intuitions arise about the best possible rate of convergence attainable by nonparametric estimators of that copula. In the end, any such estimator must be based on the marginal conditional distribution functions of the two dependent variables given the covariate, and the best possible rates for estimating such localized objects is slower than the parametric one. However, the invariance of the conditional copula given the value of the covariate suggests the possibility of parametric convergence rates. The more optimistic intuition is shown to be correct, confirming a conjecture supported by extensive Monte Carlo simulations by I. Hobaek Haff and J. Segers [Computational Statistics and Data Analysis 84:1--13, 2015] and improving upon the nonparametric rate obtained theoretically by I. Gijbels, M. Omelka and N. Veraverbeke [Scandinavian Journal of Statistics 2015, to appear]. The novelty of the proposed approach lies in the double smoothing procedure employed for the estimator of the marginal cumulative distribution functions. Under mild conditions on the bandwidth sequence, the estimator is shown to take values in a certain class of smooth functions, the class having sufficiently small entropy for empirical process arguments to work. The copula estimator itself is asymptotically undistinguishable from a kind of oracle empirical copula, making it appear as if the marginal conditional distribution functions were known.