Résumé : Astrometry covers a parameter space that cannot be reached by RV or transit methods to detect terrestrial planets on wide orbits. In addition, high accuracy astrometric measurements are necessary to measure the inclination of the planet's orbits. Here we investigate the principles of an artefact of the astrometric approach. Namely, the displacement of the photo-centre due to inhomogeneities in a dust disc around the parent star. Indeed, theory and observations show that circumstellar discs can present strong asymmetries. We model the pseudo-astrometric signal caused by these inhomogeneities, asking whether a dust clump in a disc can mimic the astrometric signal of an Earth-like planet. We show that these inhomogeneities cannot be neglected when using astrometry to find terrestrial planets. We provide the parameter space for which these inhomogeneities can affect the astrometric signals but still not be detected by mid-IR observations. We find that a small cross section of dust corresponding to a cometary mass object is enough to mimic the astrometric signal of an Earth-like planet. Astrometric observations of protoplanetary discs to search for planets can also be affected by the presence of inhomogeneities. Some further tests are given to confirm whether an observation is a real planet astrometric signal or an impostor. Eventually, we also study the case where the cross section of dust is high enough to provide a detectable IR-excess and to have a measurable photometric displacement by actual instruments such as Gaia, IRAC or GRAVITY. We suggest a new method, which consists in using astrometry to quantify asymmetries (clumpiness) in inner debris discs that cannot be otherwise resolved.