So far, no phytolith extraction protocols have been tested for accuracy and repeatability. Here we aim to display a phytolith extraction method combining the strengths of two widely used protocols, supplemented with silica microspheres as exogenous markers for quantifying phytolith concentrations. Phytolith concentrations were estimated for samples from two sedimentary sequences in which numerical age–depth models make it possible to calculate phytolith influxes (phytolith numbers per cm2per yr). Analysis of replicates confirmed the statistical robustness, the repeatability and the very few biases of our extraction technique for small phytoliths, since the relationship between grass silica short cells and microspheres was kept stable. Furthermore, we demonstrated that silica microspheres are robust exogenous markers for estimating phytolith concentrations. The minimum number of items (i.e., phytoliths plus silica microspheres) that must be counted to estimate phytolith concentrations and thus influxes depends on the ratio of phytoliths to microspheres (R) and is minimized when R = 1. Nevertheless, we recommend using ratios R ≤ 1 in order to avoid having the counting process become excessively time-consuming, because microspheres are easier to identify and count than phytoliths.