Galaxy–galaxy (g–g) lensing is a powerful probe of the relation between galaxies and dark matter (DM) haloes, but its theoretical interpretation requires a careful modelling of various contributions, such as the contributions from central and satellite galaxies. For this purpose, a phenomenological approach based on the halo model has been developed, allowing for fast exploration of the parameter space of models. In this paper, we investigate the ability of the halo model to extract information from the g–g weak lensing signal by comparing it to high-resolution dissipationless simulations that resolve subhaloes. We find that the halo model reliably determines parameters such as the host halo mass of central galaxies, the fraction of galaxies that are satellites and their radial distribution inside larger haloes. If there is a significant scatter present in the central galaxy host halo mass distribution, then the mean and median mass of that distribution can differ significantly from one another, and the halo model mass determination lies between the two. This result suggests that when analysing the data, galaxy subsamples with a narrow central galaxy halo mass distribution, such as those selected based on stellar mass, should be chosen for a simpler interpretation of the results.