The stability of arsenic species {arsenate [As(V)], monomethylarsonate [MMA], dimethylarsinate [DMA] and arsenite [As(III)]} in two types of urban wastewater samples (raw and treated) was evaluated. Water samples containing a mixture of the different arsenic species were stored in the absence of light at three different temperatures: +4 °C, +20 °C and +40 °C. At regular time intervals, arsenic species were determined by high performance liquid chromatography (HPLC)-hydride generation (HG)-atomic fluorescence spectrometry (AFS). The experimental conditions for the separation of arsenic species by HPLC and their determination by AFS were directly optimised from wastewater samples. As(III), As(V), MMA and DMA were separated on an anion exchange column using phosphate buffer (pH 6.0) as the mobile phase. Under these conditions the four arsenic species were separated in less than 10 min. The detection limits were 0.6, 0.9, 0.9 and 1.8 µg L–1 for As(III), DMA, MMA and As(V), respectively. As(V), MMA and DMA were found stable in the two types of urban wastewater samples over the 4-month period at the three different temperatures tested, while the concentration of As(III) in raw wastewater sample decreased after 2 weeks of storage. A greater stability of As(III) was found in the treated urban wastewater sample. As(III) remained unaltered in this matrix at pH 7.27 over the period studied, while at lower pH (1.6) losses of As(III) were detected after 1 month of storage. The results show that the decrease in As(III) concentration with time was accompanied by an increase in As(V) concentration.