The viability of the separation of lanthanides and uranium(VI) in the form of strongly absorbing complexes with arsenazo III (AIII) was studied with the aim to increase the sensitivity of absorbance detection in determination of these metals by capillary electrophoresis (CE). Special attention was paid to the complexation equilibria in the background electrolyte (BGE). On-capillary complexation provided better peak shapes for lanthanides compared to pre-column complexation. While the BGE composition had very little effect on the peak shape of the kinetically inert uranium(VI) complex, it played a crucial role in the peak shapes of the kinetically labile lanthanide complexes. Addition of a second ligand competing with the metallochromic ligand AIII for the metal ions was found to be critical to achieve good peak shape. The nature and concentration of the competing complexing ligand added to the BGE, the pH, and the concentration of AIII were found to exert a strong influence on the separation selectivity, peak shapes and the detection sensitivity. Several carboxylic acids were compared as BGE competing ligands and citrate provided best selectivity and peak shapes. A citrate BGE at pH 4.7 and containing 0.1 mM AIII was used for the separation of uranium(VI) (350 000 theoretical plates) and La(III) (63 000 theoretical plates) while, to separate most lanthanides and uranium(VI), a similar BGE with a lower (0.03 mM) AIII concentration was used. Using hydrostatic sampling (100 mm for 10 s) detection limits of 0.35 μM (49 ppb) La(III) and 25 μM (60 ppb) UO2 were obtained. Using on-capillary complexation, sample stacking was retained for injection times of up to at least 100 s (ca. 30-mm sample plug) without loss of peak shapes for lanthanides or recovery for La(III). When this process was used, the detection limit for La(III) was reduced to about 5 ppb. Optimal properties of metallochromic ligands for separation and detection of metals by CE are discussed.