Multiply-imaged quasars and active galactic nuclei (AGNs) observed in the mid-infrared (MIR) range are commonly assumed to be unaffected by the microlensing produced by the stars in their lensing galaxy. In this paper, we investigate the validity domain of this assumption. Indeed, that premise disregards microlensing of the accretion disc in the MIR range, and does not account for recent progress in our knowledge of the dusty torus, which has unveiled relatively compact dust emission. To simulate microlensing, we first built a simplified image of the quasar composed of (i) an accretion disc whose size is based on accretion disc theory, and (ii) a larger ring-like torus whose radius is guided by interferometric measurements in nearby AGNs. The mock quasars are created in the 10 44.2 −10 46 erg/s (unlensed) luminosity range, which is typical of known lensed quasars, and are then microlensed using an inverse ray-shooting code. We simulated the wavelength dependence of microlensing for different lensed image types and for various fractions of compact objects in the lens. This allows us to derive magnification probabilities as a function of wavelength, as well as to calculate the microlensing-induced deformation of the spectral energy distribution of the lensed images. We find that microlensing variations as large as 0.1 mag are very common at 11 µm (typically rest-frame 4 µm). The main signal comes from microlensing of the accretion disc, which may be significant even when the fraction of flux from the disc is as small as 5% of the total flux. We also show that the torus of sources with L bol 10 45 erg/s is expected to be noticeably microlensed. Microlensing may thus be used to get insight into the rest near-infrared inner structure of AGNs. Finally, we investigate whether microlensing in the mid-infrared can alter the so-called R cusp relation that links the fluxes of the lensed images triplet produced when the source lies close to a cusp macro-caustic. This relation is commonly used to identify massive (dark-matter) substructures in lensing galaxies. We find that significant deviations from R cusp may be expected, which means that microlensing can explain part of the flux ratio problem.