Metal ions drive important parts of biology, yet it remains experimentally challenging to locate their binding sites. Here we present an innovative computational approach. We use interactive steering of charged ions or small molecules in an electrostatic potential map in order to identify potential binding sites. The user interacts with a haptic device and experiences tactile feedback related to the strength of binding at a given site. The potential field is the first level of resolution used in this model. Any type of potential field can be used, implicitly taking into account various conditions such as ionic strength, dielectric constants or the presence of a membrane. At a second level, we represent the accessibility of all binding sites by modelling the shape of the target macromolecule via non-bonded van der Waals interactions between its static atomic or coarse-grained structure and the probe molecule(s). The third independent level concerns the representation of the molecular probe itself. Ion selectivity can be assessed by using multiple interacting ions as probes. This method was successfully applied to the DNase I enzyme, where we recently identified two new cation binding sites by computationally expensive extended molecular dynamics simulations.