Over the past decades, analysis of occluded carbon in phytoliths (opaline silica mineral bodies that form in and between plant cells) has become a workhorse of paleoclimate and archaeological studies. Since different plant types exhibit distinctive phytolith morphologies, their assemblages are used in identifying vegetation histories or food culture adaptations. A few direct radiocarbon AMS measurements of phytoliths have been carried out, but these measurements are difficult due to the low concentrations of phytoliths in some plant species, and the small amount of C per phytolith (<2%). In addition, no phytoliths samples of a known 14C age are available to verify measurement accuracy and precision, and to check sample preparation protocols. Background corrections are also difficult to address due to the lack of suitable material. In this work, we designed a procedure to quantify a suitable blank using SiO2 powder samples (close to the opal structure, and free of 14C). The full phytolith extraction showed high carbon contamination components: a) ∼3 μg of modern C and ∼2 μg of dead C. We also performed accuracy tests on large phytolith-occluded carbon samples extracted from soils and harvested plants. The unexpected 14C ages in some of the results triggered further investigations of possible sources of carbon contamination.