Abstract Pilot tests were conducted to investigate the removal of geosmin and 2-methylisoborneol (MIB) by new and semi-exhausted granular activated carbon (GAC) extracted from full-scale filters located in the City of Toronto's drinking water treatment facilities. Four pilot filters containing core-sampled GAC and new sand were fed with settled water from a full-scale plant and operated under conditions similar to those employed at full-scale. None of the pilot filters appeared to be capable of reducing geosmin and MIB concentrations to below the commonly cited threshold odour limits of 4 ng/L for geosmin and 9 ng/L for MIB at the influent levels tested. When operated at a 5-min empty bed contact time (EBCT) with geosmin influent concentrations in the range of about 70 to 110 ng/L, removals ranged from 10 to 38% in filters with 25 to 30 cm of used GAC. In the filter with 25 cm of new GAC, removal was 83%. When operated with a 7.5-min EBCT, the filter containing 95 cm of used bituminous GAC removed 78% of the geosmin present in the influent. For both geosmin and MIB, the effluent concentration and the amount removed increased as influent concentration increased, as was expected. In general, geosmin was better removed than MIB.