A method using accelerator mass spectrometry (AMS) has been developed to offer a more sensitive alternative to scintillation techniques for the determination of chlorine-36 (36Cl) in foods. The main problem in method development was the potential interference of the sulfur-36 (36S) isobar. This was overcome by reducing the sulfur level of acid digests of food by precipitation of chloride as silver chloride, then purification by washing, dissolution and reprecipitation to present silver chloride as the AMS target. The limit of detection was around 0.1 Bq kg−1 and the limit of quantitation was around 0.2 Bq kg−1. The AMS method was only semi-quantitative at the lowest levels of interest. To test the method a few samples of milk (five) and blackberries (three) collected from near two nuclear power stations as potential sources of contamination were analysed. Blackberries spiked at 0.2 Bq kg−1 and milk spiked at 0.1 Bq kg−1 could be distinguished from method blanks. There was no 36Cl detectable in the unspiked samples.