In this study, we introduce a two-dimensional metasurface sensor designed to detect, locate and distinguish between different objects placed in its near field. When an object is placed on the metasurface, local changes can be detected in one or more of the structure’s meta-atoms. This interaction generally modifies the inductance of the meta-atom, resulting in changes to the overall input impedance of the surface. We derive the properties of the structure and its behaviour in terms of superposition and demonstrate that observing the meta-surface from a single point is sufficient for unambiguous localisation and identification. To model these changes effectively and identify the position of an object, we employ a neural network machine learning algorithm. Our approach enables accurate localisation of all studied objects, with a precision exceeding $$98\%$$
98
%
. Additionally, the distinct signatures of the objects allow for separation between them with an accuracy of over $$97\%$$
97
%
. The potential applications of this platform extend to foreign object detection on metasurfaces for wireless power transfer, providing proximity detection for many surfaces such as clothing, car bodies and robotic carapaces. Furthermore, our research suggests the feasibility of implementing a touchscreen type interface requiring only a single waveguide connection.