(Abridged) Until recently the information inferred from gamma-ray burst follow-up observations was mostly limited to optically bright afterglows, biasing all demographic studies against sight-lines that contain large amounts of dust. Here, we present GRB afterglow and host observations for a sample of bursts that are exemplary of previously missed ones because of high visual extinction along the sight-line. This facilitates an investigation of the properties, geometry and location of the absorbing dust of these poorly-explored host galaxies, and a comparison to hosts from optically-selected samples. The hosts of the dustiest afterglows are diverse in their properties, but on average redder, more luminous and massive than the hosts of optically-bright events. We hence probe a different galaxy population, suggesting that previous host samples miss most of the massive, chemically-evolved and metal-rich members. This also indicates that the dust along the sight-line is often related to host properties, and thus probably located in the diffuse ISM or interstellar clouds and not in the immediate GRB environment. Some of the hosts in our sample, are blue, young or of small stellar mass illustrating that even apparently non-extinguished galaxies possess very dusty sight-lines due to a patchy dust distribution. The presented observations establish a population of luminous, massive and correspondingly chemically-evolved GRB hosts. This suggests that GRBs trace the global star-formation rate better than studies based on optically-selected host samples indicate, and the previously-claimed deficiency of high-mass host galaxies was at least partially a selection effect.

We present a framework to obtain photometric redshifts (photo-zs) for gamma-ray burst afterglows. Using multi-band photometry from GROND and Swift/UVOT, photo-zs are derived for five GRBs for which spectroscopic redshifts are not available. We use UV/optical/NIR data and synthetic photometry based on afterglow observations and theory to derive the photometric redshifts of GRBs and their accuracy. Taking into account the afterglow synchrotron emission properties, we investigate the application of photometry to derive redshifts in a theoretical range between z~1 and z~12. The measurement of photo-zs for GRB afterglows provides a quick, robust and reliable determination of the distance scale to the burst, particularly in those cases where spectroscopic observations in the optical/NIR range cannot be obtained. Given a sufficiently bright and mildly reddened afterglow, the relative photo-z accuracy is better than 10% between z=1.5 and z~7 and better than 5% between z=2 and z=6. We detail the approach on 5 sources without spectroscopic redshifts observed with UVOT on-board Swift and/or GROND. The distance scale to those same afterglows is measured to be $z=4.31^{+0.14}_{-0.15}$ for GRB 080825B, $z=2.13^{+0.14}_{-0.20}$ for GRB 080906, $z=3.44^{+0.15}_{-0.32}$ for GRB 081228, $z=2.03^{+0.16}_{-0.14}$ for GRB 081230 and $z=1.28^{+0.16}_{-0.15}$ for GRB 090530. Combining the response from UVOT with ground-based observatories and in particular GROND operating in the optical/NIR wavelength regime, reliable photo-zs can be obtained from z ~ 1.0 out to z ~ 10, and possibly even at higher redshifts in some favorable cases, provided that these GRBs exist, are localized quickly, have sufficiently bright afterglows and are not heavily obscured.