Different diets wear teeth in different ways and generate distinguishable wear and microwear patterns that have long been the basis of palaeodiet reconstructions. Little experimental research has been performed to study them together. Here, we show that an artificial mechanical masticator, a chewing machine, occluding real horse teeth in continuous simulated chewing (of 100 000 chewing cycles) is capable of replicating microscopic wear features and gross wear on teeth that resemble wear in specimens collected from nature. Simulating pure attrition (chewing without food) and four plant material diets of different abrasives content (at n = 5 tooth pairs per group), we detected differences in microscopic wear features by stereomicroscopy of the chewing surface in the number and quality of pits and scratches that were not always as expected. Using computed tomography scanning in one tooth per diet, absolute wear was quantified as the mean height change after the simulated chewing. Absolute wear increased with diet abrasiveness, originating from phytoliths and grit. In combination, our findings highlight that differences in actual dental tissue loss can occur at similar microwear patterns, cautioning against a direct transformation of microwear results into predictions about diet or tooth wear rate.