Energy dispersive X-ray microscopy (EDX) is a widely available, inexpensive method of characterizing the in-situ elemental composition of samples in Earth and life sciences. Common protocols and textbooks focussing on material sciences address EDX analysis of metallic samples that can be polished perfectly, whereas geoscientists often investigate specimens with prominent topography and composed of light, difcult to resolve elements. This is further compounded by the scarcity of literature surrounding the methodology of SEM–EDX in the feld of palaeontology, leading to common misinterpretations and artefacts during data acquisition. Here, the common errors in elemental composition obtained with EDX arising from surface topography and from parameters subject to user decisions are quantifed. As a model, fossil bioapatite (conodonts) and abiotic Durango apatite are used. It is shown that even microscale topography can distort measured composition by up to 34%, whereas topographic features such as tilt with respect to the electron beam lead to diferences of up to 85%. Working distance was not the most important parameter afecting the results and led to diferences in composition of up to 13%, whereas the choice of standard and its levelling with the sample surface led to inaccuracy reaching 33%. EDX results can be also afected by beam damage and the efects of acceleration voltage on sample acquisition and resolution are quantifed. An estimate is provided of the severity of errors associated with samples which cannot satisfy preparation requirements for EDX fully, such as holotypes, and with user decisions. Using a palaeontological example, recommendations are ofered for the best parameters and the relative importance of error sources are assessed. Open access in Facies