The insertion of solvent particles in molecular dynamics simulations of complex fluids is required in many situations involving open systems, but this challenging task has been scarcely explored in the literature. We propose a simple and fast algorithm (USHER) that inserts the new solvent particles at locations where the potential energy has the desired prespecified value. For instance, this value may be set equal to the system's excess energy per particle in such a way that the inserted particles are energetically indistinguishable from the other particles present. During the search for the insertion site, the USHER algorithm uses a steepest-descent iterator with a displacement whose magnitude is adapted to the local features of the energy landscape. The only adjustable parameter in the algorithm is the maximum displacement, and we show that its optimal value can be extracted from an analysis of the structure of the potential energy landscape. We present insertion tests in periodic and nonperiodic systems filled with a Lennard-Jones fluid whose density ranges from moderate to high values. (C) 2003 American Institute of Physics.