Given a curve f and a surface S, how hard is it to find a simple curve f' in S that is the most similar to f? We introduce and study this simple curve embedding problem for piecewise linear curves and surfaces in R^2 and R^3, under Hausdorff distance, weak Frechet distance, and Frechet distance as similarity measures for curves. Surprisingly, while several variants of the problem turn out to have polynomial-time solutions, we show that in R^3 the simple curve embedding problem is NP-hard under Frechet distance even if S is a plane, as well as under weak Frechet distance if S is a terrain. Additionally, these results give insight into the difficulty of computing the Frechet distance between surfaces, and they imply that the partial Frechet distance between non-planar surfaces is NP-hard as well.