I will report on the first results of a large observational program on gamma-ray burst (GRB) host galaxies (PI: Jens Hjorth), completed in 2008. The survey has targeted a carefully selected sample of 71 Swift GRBs. The bursts have been chosen primarily based on their X-ray afterglow detection, in order to avoid biases against dusty or high-redshift objects. The project has several goals. 1) The detection of the hosts through deep photometry. 2) The search for red systems through infrared photometry. 3) The determination of missing redshifts, to increase the completeness in the sample. 4) The search for Ly-alpha emission for all hosts at z > 2. 5) The study of clustering properties of short-duration GRB hosts. The main results of the program include a large host detection rate, the limited presence of red(dened) systems, the derivation of an improved (less biased) redshift distribution, a lower-than-expected Ly-alpha recovery fraction, and clues on the nature of dark GRBs.