Artificial light at night (ALAN) is an increasing phenomenon associated with worldwide urbanisation. In birds, broad-spectrum white ALAN can have disruptive effects on activity patterns, metabolism, stress response and immune function. There has been growing research on whether the use of alternative light spectra can reduce these negative effects, but surprisingly, there has been no study to determine which light spectrum birds prefer. To test such a preference, we gave urban and forest great tits (Parus major) the choice where to roost using pairwise combinations of darkness, white or green dim light at night (1.5 lux). Birds preferred to sleep under artificial light instead of darkness, and green was preferred over white light. In a subsequent experiment, we investigated the consequence of sleeping under a particular light condition, and measured birds’ daily activity levels, daily energy expenditure (DEE), oxalic acid as a biomarker for sleep debt, and cognitive abilities. White light affected activity patterns more than green light. Moreover, there was an origin-dependent response to spectral composition: in urban birds the total daily activity and night activity did not differ between white and green light, while forest birds were more active under white than green light. We also found that individuals who slept under white and green light had higher DEE. However, there were no differences in oxalic acid levels or cognitive abilities between light treatments. Thus, we argue that in naïve birds that never encountered light at night, white light might disrupt circadian rhythms more than green light. However, it is possible that negative effects of ALAN on sleep and cognition might be observed only under intensities higher than 1.5 lux. These results suggest that reducing the intensity of light pollution as well as tuning the spectrum towards long wavelengths may considerably reduce its impact.